Download REPORTS Transplacental Effects of -Azido-2

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Vectors in gene therapy wikipedia , lookup

Transcript
REPORTS
Transplacental Effects of
3*-Azido-2*,3*-Dideoxythymidine
(AZT): Tumorigenicity in Mice
and Genotoxicity in Mice
and Monkeys
Ofelia A. Olivero, Lucy M.
Anderson, Bhalchandra A.
Diwan, Diana C. Haines, Steven
W. Harbaugh, Thomas J. Moskal,
Ann B. Jones, Jerry M. Rice,
Charles W. Riggs, Daniel
Logsdon, Stuart H. Yuspa,
Miriam C. Poirier*
Background: When given during pregnancy, the drug 3*-azido-2*,3*-dideoxythymidine (AZT) substantially reduces
maternal–fetal transmission of human
immunodeficiency virus type 1 (HIV1). However, AZT has been shown to be
carcinogenic in adult mice after lifetime oral administration. In this study,
we assessed the transplacental tumorigenic and genotoxic effects of AZT in
the offspring of CD-1 mice and Erythrocebus patas monkeys given AZT
orally during pregnancy. Methods:
Pregnant mice were given daily doses
of either 12.5 or 25.0 mg AZT on days
12 through 18 of gestation (last 37% of
gestation period). Pregnant monkeys
were given a daily dose of 10.0 mg AZT
5 days a week for the last 9.5–10 weeks
of gestation (final 41%–43% of gestation period). AZT incorporation into
nuclear and mitochondrial DNA and
the length of chromosomal end (telomere) DNA were examined in multiple
tissues of newborn mice and fetal monkeys. Additional mice were followed
from birth and received no further
treatment until subjected to necropsy
and complete pathologic examination
at 1 year of age. An anti-AZT radioimmunoassay was used to monitor AZT
1602 REPORTS
incorporation into DNA. Results: At 1
year of age, the offspring of AZTtreated mice exhibited statistically significant, dose-dependent increases in
tumor incidence and tumor multiplicity
in the lungs, liver, and female reproductive organs. AZT incorporation into
nuclear and mitochondrial DNA was
detected in multiple organs of transplacentally exposed mice and monkeys.
Shorter chromosomal telomeres were
detected in liver and brain tissues from
most AZT-exposed newborn mice but
not in tissues from fetal monkeys. Conclusions: AZT is genotoxic in fetal mice
and monkeys and is a moderately
strong transplacental carcinogen in
mice examined at 1 year of age. Careful
long-term follow-up of AZT-exposed
children would seem to be appropriate.
[J Natl Cancer Inst 1997;89:1602–8]
Exposure of pregnant females of numerous mammalian species, including
nonhuman primates, to various chemical
carcinogens results in neoplasms in their
offspring (1). The human relevance of
transplacental carcinogenesis was established with the discovery that diethylstilbestrol (DES) caused vaginal adenocarcinomas in the children of women treated
during pregnancy (2). Subsequent experimental studies in mice duplicated this effect (3). Epidemiologic evidence has implicated transplacental exposures to
radiation, certain medications, pesticides,
occupational chemicals, and metals (4–6)
as possible contributors to human cancer
risk. Mechanistic studies with rodents indicate that fetuses may be particularly at
risk of tumor initiation by chemicals, with
high rates of cell division and other fetal
characteristics greatly enhancing vulnerability (7).
The nucleoside analogue 38-azido28,38-dideoxythymidine (AZT), widely
used to treat human immunodeficiency
virus type 1 (HIV-1) infection, has become the standard of care in preventing
fetal transmission of the virus in HIV-1positive pregnant women (8,9). Recom-
mended treatment encompasses five daily
100.0-mg doses (approximately 8.3 mg/
kg body weight per day) during the second and third trimesters of pregnancy,
with additional maternal intravenous dosing at delivery and oral AZT given to the
infant after birth (9,10). In a recent study
from the Pediatric AIDS (i.e., acquired
immunodeficiency syndrome) Clinical
Trials Group (9), this regimen reduced viral transmission to the fetus from 22.6%
(n 4 204) to 7.6% (n 4 198).
In adult mice, AZT is carcinogenic.
Ayers et al. (11) reported 10% and the
National Toxicology Program (12) reported 22% incidences of vaginal squamous papillomas and carcinomas after
lifetime oral administration of the drug to
CD-1 and B6C3F1 mice, respectively. At
similar doses in mice exposed to AZT for
28 days, dose-related incorporation of
AZT into vaginal DNA and increased
vaginal epithelial proliferation were observed (13). Because of the projected
widespread use of AZT in human pregnancy, we have investigated the genotoxic
and carcinogenic effects of AZT in the
*Affiliations of authors: O. A. Olivero, S. H.
Yuspa, M. C. Poirier, Laboratory of Cellular Carcinogenesis and Tumor Promotion, Division of Basic
Sciences, National Cancer Institute, Bethesda, MD;
L. M. Anderson, A. B. Jones, Laboratory of Comparative Carcinogenesis, Division of Basic Sciences,
National Cancer Institute, National Cancer InstituteFrederick Cancer Research and Development Center, Frederick, MD; B. A. Diwan (Intramural Research Supported Program), D. C. Haines (Pathology/
Histotechnology Laboratory), D. Logsdon (Laboratory Animal Sciences Program), Science Applications International Corp.–National Cancer InstituteFrederick Cancer Research and Development
Center; S. W. Harbaugh, T. J. Moskal, BioQual,
Inc., Rockville, MD; J. M. Rice, International
Agency for Research on Cancer, Lyon, France; C.
W. Riggs, Data Management Services, Inc., National Cancer Institute-Frederick Cancer Research
and Development Center.
Correspondence to: Ofelia A. Olivero, Ph.D., National Institutes of Health, Bldg. 37, Rm. 3B15,
MSC-4255, Bethesda, MD 20892–4255. E-mail:
[email protected]
See ‘‘Notes’’ following ‘‘References.’’
© Oxford University Press
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
offspring of exposed pregnant CD-1 mice
and AZT genotoxicity in the fetuses of
exposed pregnant Erythrocebus patas
monkeys.
Materials and Methods
Source and Purity of AZT
AZT (lot No. 063H7819) was obtained from
Sigma Chemical Co., St. Louis, MO. The purity of
AZT was assessed by means of elemental analysis,
UV spectrum analysis, nuclear magnetic resonance
spectrometry, and mass spectrometry, as well as by
means of high-performance liquid chromatography
with two different mobile phases (data not shown).
All of the parameters were consistent with literature
values, confirming the AZT structure and assessing
the purity of this lot at greater than 99.8%. The
determinations were performed by Dr. Haleem J.
Issaq of the Chemical Synthesis and Analysis Laboratory, Science Applications International Corp.
(SAIC), National Cancer Institute (NCI)-Frederick
Cancer Research and Development Center
(FCRDC), Frederick, MD.
Tumorigenicity in the Offspring of
Pregnant Mice Receiving AZT
Animal care for this experiment was provided in
accordance with the procedures outlined in the
Guide for the Care and Use of Laboratory Animals
[National Institutes of Health publication No.
(NIH)86–23, 1985] under an animal study proposal
approved by the FCRDC Animal Care and Use
Committee. Female CD-1 Swiss mice (Charles
River, Raleigh, NC) were given food and water ad
libitum. The mice were mated, and day 1 of pregnancy was ascertained by the presence of a vaginal
plug. Forty-five pregnant mice were given 0 (17
litters), 12.5 (13 litters), or 25.0 (15 litters) mg AZT
in 0.5 mL sterile distilled water via an intragastric
tube once daily on days 12 through 18 of gestation.
Preliminary toxicity studies indicated that the maximum fetal-tolerated daily dose was 25.0 mg (approximately 420 mg/kg term body weight), with
doses of 30.0–50.0 mg causing fetal or newborn
loss. The pups were delivered normally. The average
number of pups weaned per litter was 11.7, 12.1, and
10.1 for the control, low-dose, and high-dose
groups, respectively. Ten mice per sex from each
group of pups were killed on schedule at 13, 26, and
52 weeks after delivery. No tumors were evident at
the first two time points. However, at 52 weeks, lung
and liver tumors were observed in the AZT-exposed
mice, so additional animals (selected at random)
were killed to obtain numbers adequate for statistical
analysis. In addition, moribund mice from all groups
were killed, and the findings from these mice were
included with the findings from the mice killed at 52
weeks. Altogether, the numbers of litters represented
were five (61 mice) for the control group (i.e., no
AZT), six (45 mice) for the 12.5-mg-AZT group,
and nine (50 mice) for the 25.0-mg-AZT group. All
remaining mice will be killed at 2 years of age.
When moribund or at planned sacrifice, mice underwent a complete necropsy. Lungs, liver, ovary/
testis, uterus/cervix/vagina, thymus, spleen, lymph
nodes, kidneys, brain, pituitary, mammary gland, femur, cecum, and all grossly noted lesions and
masses were examined by means of light microscopy. The pathology findings were peer-reviewed
by Dr. Miriam Anver, SAIC, NCI-FCRDC, and by
Dr. Jerrold Ward, Office of Laboratory of Animal
Sciences, NCI. All lung, liver, and female reproductive tract lesions were reviewed, as well as all questionable and representative definitive hematopoietic
lesions. Spiral organisms, thought to be Helicobacter hepaticus, were observed in Steiner’s stained
slides of most ceca, but only a few organisms were
observed in one liver, and no hepatitis was observed
in any of the livers. The mice were otherwise specific pathogen free. Tumor data are presented as
tumor incidence (percent of animals with tumors)
and tumor multiplicity (number of tumors per animal).
Statistical Methods
Dose-dependent linear trends of tumor incidence
proportions were evaluated by use of the Cochran–
Armitage chi-squared test (two-tailed) (14), and
high-dose versus control comparisons were made by
use of Fisher’s exact test (two-tailed) (15). Twotailed chi-squared tests (16) were used to analyze the
reduction in hematopoietic neoplasms in AZTexposed males and females in relationship to control
animals. Tests of the homogeneity of the proportions
of tumor bearers per litter within each of the sex–
dose–organ groups were performed by use of the
correlated binomial C(a) test statistic as described
by Tarone (17). In addition, for lungs and liver, comparisons of low-dose treatment with high-dose treatment on the basis of litters involved the analysis of
variance of 1) raw proportions of tumor-bearing animals in each litter and 2) Freeman–Tukey arcsine
transformations of the proportions (18), with each
analysis using weights proportional to the litter
sizes. Tumor multiplicities were tested for dosedependent trends by use of the nonparametric Jonckheere test (two-tailed) (19), and high-dose versus
control comparisons were made by use of the nonparametric Wilcoxon rank-sum test (two-tailed)
(20).
Incorporation of AZT Into Nuclear
and Mitochondrial DNA of Fetal
CD-1 Mice Exposed In Utero to AZT
Pregnant CD-1 mice were given 0 (n 4 3) or 25.0
(n 4 3) mg AZT/day by gavage on days 12 through
18 of gestation (final 37% of gestation period), and
the pups (14 each for litters 1 and 3 and 11 for litter
2) were born 24 hours or less after the last exposure.
For each organ (e.g., kidneys), tissue from all of the
pups of one litter was combined. The samples were
homogenized, and DNA was isolated by means of
the Oncor nonorganic extraction method (Oncor,
Inc., Gaithersburg, MD). Mitochondrial DNA was
prepared as previously described (21).
DNA samples, in distilled water, were sonicated
for 30 seconds and then boiled for 5 minutes. Threemicrogram aliquots of DNA were subsequently assayed by use of a competitive anti-AZT radioimmunoassay (AZT-RIA) as previously described (22).
Briefly, a rabbit polyclonal anti-AZT antibody
(Sigma Chemical Co.) that also recognizes AZT in
DNA was reconstituted in 20 mL 10 mM Tris buffer
(pH 8.0), representing a 1:5000 dilution of the antibody. A 0.1-mL aliquot of the diluted antibody was
incubated in a 12 × 75-mm disposable glass tube
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
with an equal volume of a solution containing either
standard AZT plus 3 mg calf thymus carrier DNA
(Sigma Chemical Co.) or a 3-mg sample of DNA
from AZT-treated or untreated (control) animals for
90 minutes at 37 °C. Approximately 20 000 cpm
[3H]AZT tracer (20 Ci/mmol; Moravek Biochemicals, Inc., Mountain View, CA), in a volume of 100
mL, was added per tube together with 100 mL goat
anti-rabbit immunoglobulin G secondary antibody
(ICN Biomedicals, Inc., Costa Mesa, CA; reconstituted in 12 mL 10 mM Tris buffer [pH 8.0]), and the
mixture was incubated for 25 minutes at 4 °C. The
mixture was centrifuged at 2000g for 15 minutes at
4 °C. The resulting supernatant was decanted, and
the pellets were dissolved in 100 mM NaOH and
counted in a liquid scintillation counter.
The AZT-RIA employs a radiolabeled AZT
tracer, used in a constant amount in each tube, to
compete with the AZT standard, used at different
concentrations, or sample DNA for binding to the
anti-AZT antibody. In the absence of standard AZT
or sample DNA containing AZT, the highest level of
radiolabel binding to the antibody will be obtained
(cpm with no inhibitor). Because of the competitive
nature of this assay, the added presence of nonradiolabeled AZT (i.e., the standard or in the sample
DNA) will result in the inhibition of tracer-antibody
binding (cpm with inhibitor). The inhibition of
tracer-antibody binding observed with a particular
amount (‘‘x’’) of unlabeled AZT is expressed as a
percent according to the following formula:
Percent inhibition =
[(cpm with no inhibitor)
− (cpm with amount ‘‘x’’ of inhibitor)]
× 100.
(cpm with no inhibitor)
The amount of standard AZT added to 3 mg carrier
calf thymus DNA that inhibited antibody binding by
50% was 570 ± 420 fmol (mean ± standard deviation
[SD]; n 4 12) for the mouse assays. The amount of
AZT in 3 mg of biologic sample DNA was obtained
by comparing DNA from the corresponding tissue of
an untreated animal with DNA from a treated animal, calculating the percent inhibition of antibody
binding, and reading the amount of AZT from a
plotted standard curve. Each sample was assayed in
three to five separate radioimmunoassays.
Incorporation of AZT Into Fetal
Erythrocebus patas Monkey Nuclear
and Mitochondrial DNA
Monkeys were maintained and treated under
American Association for Accreditation of Laboratory Animal Care-approved conditions at BioQual,
Inc. (Rockville, MD), in accordance with humane
principles for laboratory animal care. Protocols were
reviewed and approved by the Animal Care and Use
Committee of BioQual, Inc. Pregnancies were ascertained as described (23). Pregnant patas monkeys
were given 0 (n 4 3) or 10.0 (n 4 3) mg of AZT/
day (approximately 1.5 mg/kg body weight per day)
in a piece of banana, 5 days per week for the last
9.5–10 weeks of gestation (final 41%–43% of gestation period). Dose ingestion was confirmed by an
observer. Fetuses were taken by cesarean section,
performed on the monkeys under Telazol and isofluorane anesthesia, on days 149 through 152 of gestation (24 hours after the last dose). Tissue processing for DNA preparation and the AZT-RIA were as
REPORTS 1603
described above for newborn mouse tissues, except
that the 50% inhibition value for the RIA was 820 ±
290 fmol AZT (mean ± SD; n 4 25).
Examination of Telomere Length in
Mouse and Monkey Tissues
The length of telomeric (chromosomal end) DNA
(24) was examined in DNA from the tissues of 10
newborn mouse litters, either unexposed (n 4 5) or
exposed (n 4 5) in utero to 25.0 mg AZT/day on
days 12 through 18 of gestation. For each litter, tissues from different organs were combined and homogenized with a Dounce homogenizer, and highmolecular-weight DNA was prepared by use of the
nonorganic extraction method of Oncor, Inc. The
DNA was digested with Alu I, Rsa I, and Sau 3A I
restriction endonucleases (New England Biolabs,
Inc., Beverly, MA) and resolved in 1% agarose gels
along with biotinylated molecular markers (Oncor,
Inc.). The DNA was transferred to nylon support
membranes overnight by means of capillary action
(25), and it was cross-linked to the membranes by
use of UV light. The membranes were blocked for
30 minutes at 45 °C with blocking solution (Oncor,
Inc.) and hybridized with a biotinylated human telomeric repeat sequence probe (Oncor, Inc.) overnight
in a sealed bag at 45 °C. Posthybridization washes
were performed in a solution of 0.16× standard saline citrate and 0.1% sodium dodecyl sulfate for 1
hour at 60 °C. The membranes were subsequently
blocked with a 5% low-fat milk solution at room
temperature for 30 minutes, and an amplified alkaline phosphatase immunoblot assay kit (Biorad
Laboratories, Hercules, CA) was used in membrane
staining. The sizes of the telomere repeats were determined by comparison with the biotinylated molecular markers. A similar approach was used to
examine telomere length in DNA from the tissues of
six fetal monkeys exposed to either 0 (n 4 3) or
10.0 mg AZT/day for the last 9.5–10 weeks of gestation (n 4 3).
fold, dose-dependent increase in the incidence (Fig. 1) and the multiplicity (Table
1) of tumors in the AZT-exposed groups.
The spontaneous liver and lung adenomas in the control animals are typical for
this strain of mice (26). At the same sites,
incidences of these types of tumors were
increased twofold to eightfold in AZTtreated animals (Fig. 1, A–C). For both
sexes combined, the incidence of lung
carcinomas increased significantly with
AZT dose (3%, 7%, and 14% for the 0-,
12.5-, and 25.0-mg AZT groups, respectively; P 4 .037, two-tailed chi-squared
test). Neoplasms of the female organs
(ovary, uterus, and vagina) were absent
from the control animals but were detected in 14% and 17% of the AZTexposed female offspring at the low and
high doses, respectively (Fig. 1, D).
Neoplasms of the hematopoietic system, including lymphomas, myelogenous
leukemia, and histiocytic sarcoma, which
occur spontaneously in this mouse strain,
were reduced from 33% and 16% in unexposed females and males, respectively,
to 8% in the mouse pups of both sexes
(Fig. 1, E and F). This reduction suggests
that an endogenous oncogenic retrovirus
may be inhibited by the AZT treatment
and/or that alterations in the immune system may occur in the offspring exposed to
AZT in utero.
Statistical Analyses for the
Tumor Study
Probability values for the tumor incidences presented in Fig. 1 are shown in
the legend, and those for the tumor mul-
Results
Evaluation of AZT as a
Transplacental Carcinogen
To test for transplacental tumorigenicity, pregnant CD-1 Swiss dams were exposed once daily to 0, 12.5, or 25.0 mg
AZT via intragastric intubation on days
12 through 18 of gestation. In the carcinogenicity study, litter sizes did not differ
substantially between AZT-exposed and
control mice, and no further exposures
were given postnatally. Ten offspring of
each sex from each treatment group were
necropsied at 3 and 6 months, and no neoplasms were found. However, the interim
sacrifice at approximately 1 year revealed
the presence of tumors, prompting the
sacrifice of additional animals of each sex
in each group for further analysis. Complete necropsy and histopathologic analysis of dead, moribund, or sacrificed animals at 190–400 days of age (average
ages of 350–382 days) revealed a several1604 REPORTS
Fig. 1. Tumor incidences in internal organs of offspring of pregnant mice 190–400 days after transplacental
38-azido-28,38-dideoxythymidine (AZT) exposure. Pregnant CD-1 mice were given 0, 12.5, or 25.0 mg AZT
in sterile distilled water once daily on days 12 through 18 of gestation. Pups were delivered normally and
given no further treatment. When moribund or at sacrifice (190–400 days), the mice underwent complete
necropsy (see ‘‘Materials and Methods’’ section). The figure shows the following tumor incidences: lung
tumors (from bronchoalveolar cells) in males (A) and in females (B); hepatocellular adenomas, including two
carcinomas, in males (C); female reproductive organ tumors (D); and hematopoietic tumors in males (E) and
females (F). Tumor incidences were analyzed for dose-related linear trend by use of the Cochran–Armitage
chi-squared test (two-sided). High-dose AZT (25.0 mg AZT) versus control (no AZT) comparisons employed the Fisher’s exact test (two-sided). Probability values for these two tests are, respectively, .0004 and
.00066 for male lung tumors (A); .037 and .056 for female lung tumors (B); .0011 and .0014 for male liver
tumors (C); and .033 and .034 for female reproductive organ tumors (D).
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
tiplicities in Table 1 are shown in a table
footnote. The presented statistical relationships are based on pooled groups,
since preliminary tests indicated that litter
proportions within each test group were
generally homogeneous. A possible exception was noted for the male–lung–
high-dose group, where near-significant
heterogeneity was found (P 4 .058),
which was due exclusively to one litter
with lung tumors in all five male offspring. Reanalysis of the data excluding
this litter resulted in somewhat lower significance probabilities but no overall
change in conclusions. When analyzed on
a litter basis, comparisons of low- with
high-dose AZT for the lungs and the liver
produced results that generally reaffirmed
the conclusions based on pooled groups.
Transplacental AZT Exposure and
Genotoxicity in Newborn Mice and
Fetal Monkeys
Using a sensitive anti-AZT radioimmunoassay (AZT-RIA) (22) and comparing DNA from exposed and unexposed
animals (see ‘‘Materials and Methods’’
section), we detected AZT incorporation
in the nuclear and mitochondrial DNA
from multiple pooled organs of newborn
CD-1 mouse pups (three litters) after
transplacental exposure to 25.0 mg AZT/
day on days 12 through 18 of gestation
(Table 2, A). Incorporation was widely
variable among litters, organs, and DNA
compartments, suggesting that undefined
pharmacokinetic and specific tissue factors influence persistent AZT incorporation. Litter size appeared to influence
measurable incorporation levels, since litters 1 and 3 each consisted of 14 pups and
had the lowest incorporation levels, and
litter 2, with only 11 pups, had higher
incorporation levels.
Telomerase is active in fetal tissues
and tumor cells (27), and previous studies
have demonstrated that AZT can be preferentially incorporated into the telomeric
DNA of cells containing telomerase (24).
In vitro long-term exposure to 800 mM
AZT was shown to produce an irreversible shortening of telomeres (28). In this
study, telomere length was examined in
five litters of newborn mice exposed to
25.0 mg AZT/day as described in Table 2,
A. Fig. 2 shows the results for liver,
lungs, and brain from one litter. In com-
parison with the DNA from control animals, smaller sized telomeric DNA was
detected in the livers from five litters, the
brains from three of five litters, the lungs
from two of five litters, and the kidneys
from one of five litters of AZT-treated
animals. The variability between litters
observed for telomere length is similar to
that seen for AZT incorporation into
DNA.
The data in mice demonstrated AZTinduced genotoxic and carcinogenic effects at a high dose. To approximate the
human exposure of about 8.3 mg/kg body
weight per day, we gave three pregnant
Erythrocebus patas monkeys 10.0 mg
AZT/day (approximately 1.5 mg/kg body
weight per day) 5 days a week during the
last 9.5–10 weeks of a 23-week gestation.
Multiple fetal tissues were obtained after
cesarean section. AZT incorporation was
observed in the nuclear and mitochondrial
DNA of transplacentally exposed monkeys but not in the DNA of unexposed
control animals (Table 2, B). Even though
the monkey dose was much lower than
the dose received by the mouse, incorporation levels were generally several-fold
higher than those observed in the mouse.
Table 1. Tumor multiplicities at 190–400 days of age in offspring of mice given 0, 12.5, or 25.0 mg 38-azido-28,38-dideoxythymidine (AZT) per day by
gavage on days 12 through 18 of gestation*
Organ of tumorigenesis
AZT exposure, mg
Offspring
evaluated
0
31 males
0.10 ± 0.07 (1 CA)
0.23 ± 0.14
30 females
0.13 ± 0.06 (1 CA)
0
0
23 males
0.13 ± 0.07 (1 CA)
0.48 ± 0.18 (1 CA)
na
22 females
0.14 ± 0.07 (2 CA)
0
2 uterine endometrial
stromal polyps; 1
vaginal
leiomyosarcoma
26 males
24 females
0.50 ± 0.11 (4 CA)
0.38 ± 0.10 (3 CA)
0.79 ± 0.26 (1 CA)
0
na
1 Sertoli cell tumor,
ovary; 1 histiocytic
sarcoma, uterus; 1
hemangiosarcoma,
uterus; 1 endometrial
stromal polyp, uterus
12.5
25.0
Lungs, mean No. of
tumors ± SE†
Liver, mean No. of
tumors ± SE‡
Female organs, No. of
tumors
na
Incidental, No. of tumors
2 liver
hemangiosarcomas
1 skin basal cell tumor
1 skin papilloma; 1 skin
hemangiosarcoma
None
None
1 skin papilloma; 1
malignant tumor§
*Tumor multiplicity refers to the number of tumors per animal. Results include findings from moribund and dead animals and those killed at 12–13 months. SE
4 standard error; CA 4 carcinoma; na 4 not applicable. P values (two-tailed) for trends and comparisons were as follows: Male lung tumors—trend test, P 4
.014; control (0 mg AZT) versus high-dose AZT (25.0 mg), P 4 .001; low-dose AZT (12.5 mg) versus high-dose AZT, P 4 .012. Female lung tumors—trend test,
P 4 .15; control versus high-dose AZT, P 4 .042; low-dose AZT versus high-dose AZT, P 4 .071. Male liver tumors—trend test, P 4 .013; control versus
high-dose AZT, P 4 .0016; control versus low-dose AZT, P 4 .097. Trend tests were performed with the Jonckheere test, and pairwise treatment comparisons were
made by use of the Wilcoxon rank-sum test (both are nonparametric methods).
†Lung alveolar cell adenomas and carcinomas.
‡Liver hepatocellular adenomas.
§Autolyzed neoplasm involving pancreas and spleen.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
REPORTS 1605
Table 2, A. Incorporation of 38-azido-28,38-dideoxythymidine (AZT) into nuclear and mitochondrial
DNA of fetal CD-1 mice exposed in utero to AZT*
fmol AZT/mg DNA†
Litter 1
Organ
Brain
Lungs
Liver
Kidneys
Skin
Litter 2
Litter 3
Nuclear
Mt
Nuclear
Mt
Nuclear
Mt
nd
nd
nd
23.2 ± 2.5
nd
nd
nd
nd
28.4 ± 7.5
160.8 ± 43.6
nd
226.7 ± 128.4
121.1 ± 78.9‡
41.2 ± 1.2‡
87.5 ± 23.5‡
49.5 ± 3.9‡
nd
42.9 ± 14.9‡
nd
302.7 ± 64.0‡
nd
40.3 ± 9.4
nd
nd
ns
nd
51.4 ± 15.0
nd
33.3 ± 13.3‡
227.0 ± 195.0‡
*Pregnant CD-1 mice were given 0 (n 4 3) or 25.0 (n 4 3) mg AZT/day, and the pups (14 each for litters
1 and 3 and 11 for litter 2) were born 24 hours or less after the last exposure. For each organ (e.g., kidneys),
tissue from all of the pups of one litter was combined, nuclear and mitochondrial DNA was prepared, and
3-mg aliquots of DNA were assayed by use of an anti-AZT radioimmunoassay (see ‘‘Materials and Methods’’ section). Incorporation is expressed as fmol AZT/mg DNA in relation to results obtained with tissue
DNA from unexposed mice. Except where noted, values are means ± standard error for three to five assays.
†Mt 4 mitochondrial; nd 4 not detectable; ns 4 no sample.
‡Mean ± range for samples assayed twice.
Table 2, B. Incorporation of AZT into fetal Erythrocebus patas monkey nuclear and
mitochondrial DNA*
fmol AZT/mg DNA†
Monkey No. R105
Organ
Brain
Cerebellum
Lungs
Liver
Kidneys
Heart
Placenta
Monkey No. R200
Monkey No. R226
Nuclear
Mt
Nuclear
Mt
Nuclear
Mt
nd
227.7 ± 94.7
739.0 ± 221.1
566.0 ± 274.0
703.0 ± 262.6
541.6 ± 84.9
nd
120.0 ± 40.0‡
nd
253.7 ± 135.9
nd
nd
155.0 ± 63.7
22.0 ± 6.0‡
150.0 ± 70.0‡
81.5 ± 11.5‡
nd
230.0 ± 70.0‡
170.0 ± 30.0‡
60.0 ± 10.0‡
nd
nd
ns
nd
55.0 ± 15.0‡
55.0 ± 5.0‡
45.0 ± 5.0‡
60.0 ± 0‡
nd
252.7 ± 84.0
166.7 ± 13.3
146.0 ± 69.1
316.6 ± 101.1
316.0 ± 152.8
341.7 ± 123.8
242.0 ± 89.4
282.0 ± 54.7
203.0 ± 8.8
nd
nd
324.3 ± 42.4
nd
*Pregnant patas monkeys were given 0 (n 4 3) or 10.0 (n 4 3) mg of AZT/day (approximately 1.5 mg/kg
body weight per day) in a piece of banana 5 days per week for the last 9.5–10 weeks of gestation. Fetuses
were taken by cesarean section on days 149 through 152 of gestation (24 hours after the last dose). Tissue
processing and AZT-DNA radioimmunoassay were as described in the ‘‘Materials and Methods’’ section,
with comparison of DNAs from unexposed and exposed monkeys. Except where noted, values are means ±
standard error for three to five assays.
†Mt 4 mitochondrial; nd 4 not detectable; ns 4 no sample.
‡Mean ± range of two assays.
Fig. 2. Telomere length in nuclear DNA from
lung, brain, and liver tissue of newborn mice
either unexposed (c, control) or exposed (a) to
38-azido-28,38-dideoxythymidine (AZT) in utero
at a dose of 25.0 mg/day on days 12 through 18
of gestation. DNA was isolated, digested with
restriction endonucleases, resolved in a 1% agarose gel, transferred to a nylon support membrane, and hybridized with a biotinylated probe
specific for human telomeric DNA repeat sequences. The sizes of the telomeric repeats were
determined by comparison with the biotinylated
molecular marker (MM) DNAs (note the 2- and
23-kilobase reference regions). See ‘‘Materials
and Methods’’ section for more details.
Again, there was a large interanimal variability in AZT-DNA levels among the
monkey fetuses and in the DNA compartments. The size of telomeric DNA was
1606 REPORTS
not measurably altered in the DNA from
liver, brain, cerebellum, heart, lungs, and
placenta obtained from AZT-exposed fetal monkeys (data not shown).
Discussion
At the doses tested here, AZT is unequivocally a transplacental genotoxin
and carcinogen in CD-1 mice. The liver
and lungs, targets for tumor formation in
this study, are typical organ sites for
genotoxic transplacental carcinogens in
mice (7). The incidence, latency, multiplicity, and histopathology of the AZTinduced tumors indicate that AZT is
intermediate in potency as a mouse transplacental carcinogen. On a toxic-equivalent dose basis, this drug is less potent
than N-nitrosoethylurea (29), 7,12dimethylbenz[a]anthracene (30), and 3methylcholanthrene (31), but it is more
potent than N-nitrosodimethylamine (29)
and the tobacco-specific nitrosamine 4(methylnitrosamino)-1-(3-pyridyl)-1butanone (32). When given transplacentally to mice, benzo[a]pyrene produced
lung and liver tumor multiplicities similar
to those observed here (30). The female
reproductive organ tumors, absent in untreated animals, are similar in type and
incidence to those resulting from mouse
transplacental exposure to DES (3).
In these experiments, AZT was given
to mice for the last 37% of gestation at a
daily dose that was approximately fivefold higher than the equivalent daily dose
received by pregnant women [with the
use of the mouse–human scaling factor of
1:12 described by Freireich et al. (33)].
With no scaling factors applied, an HIV1-positive pregnant woman receiving
AZT during the last two trimesters will
have a total dose of about 1.4 g/kg body
weight. The mice receiving 25.0 mg AZT/
day were given a total dose of about 3.5
g/kg body weight, and the monkeys in this
study received about 0.08 g AZT/kg body
weight. Nevertheless, fetal Erythrocebus
patas monkeys, given AZT at a much
lower daily dose than that received by the
CD-1 mice, had more AZT-DNA incorporation than that detected in the newborn
mice.
Human–mouse dose comparisons are
further complicated by a number of factors. Phosphorylation of AZT is more efficient in mice than in humans (34). The
plasma half-life of AZT in mice is 20
minutes (35,36), whereas it is 1–2 hours
in humans (10). Because CD-1 mice typically carry 10–14 pups, the ratio of fetal
weight to total maternal weight at deliv-
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
ery is 0.4 or more in the mouse, compared
with 0.1 or less in humans. Together with
tissue-specific differences in target organ
responses and differences in litter size
(11–14 pups for these experiments), the
variability in AZT-DNA incorporation
and telomere length might be expected.
However, despite this variability, litter
differences in tumorigenesis were not
significant (Fig. 1). Pregnant monkeys,
which carry only one fetus, may therefore
be a more appropriate model for the
human.
Two transplacental studies of AZT in
mice, in which much lower doses than our
doses of 12.5 and 25.0 mg/day were used,
failed to find a carcinogenic effect. Ayers
et al. (37) gave pregnant CD-1 mice 20.0
and 40.0 mg AZT/kg body weight per day
(approximately 0.5 and 1.0 mg AZT/day)
from the 10th day of gestation through
weaning. Pups were then treated with 0,
20.0, or 40.0 mg AZT/kg body weight per
day for 24 months. In animals given lifetime exposure to 40.0 mg/kg body weight
per day, the males were unaffected, and
vaginal tumors were observed only in the
adult females; however, the tumor yield
was not increased in animals given no
drug exposure past weaning. Bilello et al.
(35) gave 4.5 mg AZT/day to pregnant
mice on days 16 through 21 of gestation
and postnatally to nursing dams; these investigators did not detect tumors in the
offspring by gross examination of tissues
at 18 months. Taken together with our
study, the data suggest that cumulative
dose effects are critically important in
determining the prenatal carcinogenicity
of AZT.
Two novel genotoxicity assays were
employed in this study. The AZT-RIA,
which has previously been validated by
comparison with incorporation of radiolabeled AZT into nuclear DNA (22), was
applied here for the first time to mitochondrial DNA. Unlike the situation with
many chemical carcinogens that preferentially modify mitochondrial DNA by covalent binding, the incorporation of AZT
into mitochondrial DNA was highly variable and did not parallel the incorporation
of the drug into nuclear DNA. Because
there is no literature for the comparison of
nuclear and mitochondrial DNA incorporation of AZT, further validation must
await confirmation and the development
of alternative methods. Although it is unclear whether mitochondrial genotoxicity
is related to tumorigenicity, it may be significant that mice developed tumors in organs where nuclear AZT DNA was not
always detectable. In addition, the monkeys had higher levels of AZT incorporation into DNA than the mice in spite of
receiving a much lower dose. The telomere length assays also provide novel information but no clear context for evaluation of the biologic consequences. The
observation that telomeres were shortened
in the mice and not in the monkeys would
suggest that dose effects are important;
however, the fetal mice appeared to function normally, and any possible relationship with the process of tumorigenesis remains obscure. Both of these biomarkers
will be examined in great detail in future
studies to assess their relationships with
the observable biologic consequences of
AZT exposure.
The relevance of the mouse studies to
human exposure must be considered in
the context of dose-equivalency, an especially difficult extrapolation for transplacental exposures. Available literature
does not allow an accurate estimation of
human risk implied by these data. However, our results suggest that the current
practice of treating HIV-1-positive
women and their infants with high doses
of AZT could increase cancer risk in the
drug-exposed children when they reach
young adulthood or middle age. The remarkable effectiveness of AZT in preventing fetal HIV infection (8,9) indicates
that the immediate need for treatment of a
potentially fatal disease should outweigh
the potential cancer risk. However, given
the relatively high tumor incidences observed here at only half of the lifetime of
the mouse, it would seem appropriate
both to include additional notification in
informed consent documents and to plan
extensive follow-up of AZT-exposed
children. In addition, since human carcinogenesis is multifactorial and takes
many years to develop (38), protective
modulation may be possible (39).
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
References
(1) Napalkov MP, Rice JM, Tomatis L, Yamasaki
H, editors. Perinatal and multigeneration carcinogenesis. Lyon, France: IARC, 1989.
(2) Herbst AL, Ulfelder H, Poskanzer DC. Adenocarcinoma of the vagina. Association of maternal stilbestrol therapy with tumor appearance
in young women. N Engl J Med 1971;284:
878–81.
(3) Newbold R. Cellular and molecular effects of
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
(16)
(17)
(18)
developmental exposure to diethylstilbestrol:
implications for other environmental estrogens.
Environ Health Perspect 1995;103 Suppl 7:
83–7.
Zahm SH, Devesa SS. Childhood cancer: overview of incidence trends and environmental
carcinogens. Environ Health Perspect 1995;
103 Suppl 6:177–84.
Savitz DA, Chen JH. Parental occupation and
childhood cancer: review of epidemiologic
studies. Environ Health Perspect 1990;88:
325–37.
Bunin GR, Rose PG, Noller KL, Smith E. Occupational and environmental reproductive
hazards. In: Paul M, editor. A guide for clinicians. Baltimore (MD): Williams & Wilkins,
1993:76–88.
Anderson LM, Donovan PJ, Rice JM. Risk assessment for transplacental carcinogenesis. In:
Li AP, editor. New approaches in toxicity testing and their application in human risk assessment. New York: Raven Press, 1985:
179–202.
Connor EM, Sperling RS, Gelber R, Kiselev P,
Scott G, O’Sullivan MJ, et al. Reduction of
maternal–infant transmission of human immunodeficiency virus type 1 with zidovudine
treatment. Pediatric AIDS Clinical Trials
Group Protocol 076 Study Group. N Engl J
Med 1994;331:1173–80.
Sperling RS, Shapiro DE, Coombs RW, Todd
JA, Herman SA, McSherry GD, et al. Maternal
viral load, zidovudine treatment, and the risk of
transmission of human immunodeficiency virus type 1 from mother to infant. Pediatric
AIDS Clinical Trials Group Protocol 076
Study Group. N Engl J Med 1996;335:
1621–9.
Retrovir. In: Physicians’ desk reference. Montvale (NJ): Medical Economics Co., 1997:
1216–25.
Ayers KM, Clive D, Tucker WE Jr, Hajian G,
De Miranda P. Nonclinical toxicology studies
with zidovudine: genetic toxicity tests and carcinogenicity bioassays in mice and rats. Fundam Appl Toxicol 1996;32:148–58.
Toxicology and carcinogenesis studies of AZT.
Research Triange Park (NC): National Toxicology Program, 1996.
Olivero OA, Beland FA, Fullerton NF, Poirier
MC. Vaginal epithelial DNA damage and expression of preneoplastic markers in mice during chronic dosing with tumorigenic levels of
38-azido-28,38-dideoxythymidine. Cancer Res
1994;54:6235–42.
Armitage P. Statistical methods in medical research. Oxford, England: Blackwell Scientific
Publications, 1971:362–7.
Colquhoun D. Lectures on biostatistics. Oxford, England: Clarendon Press, 1971:
116–23.
Snedecor GW, Cochran WG. Statistical methods, sixth edition. Ames (IA): Iowa State University Press, 1967:215–21.
Tarone RE. Testing the goodness of fit of the
binomial distribution. Biometrika 1979;66:
585–90.
Freeman MF, Tukey JW. Transformations related to the angular and the square root. Ann
Math Statist 1950;21:607–11.
REPORTS 1607
(19) Hollander M, Wolfe DA. Nonparametric statistical methods. New York: John Wiley and
Sons, 1973:120–3.
(20) Hollander M, Wolfe DA. Nonparametric statistical methods. New York: John Wiley and
Sons, 1973:67–74.
(21) Tamura K, Aotsuka T. Rapid isolation method
of animal mitochondrial DNA by the alkaline
lysis procedure. Biochem Genet 1988;26:
815–9.
(22) Olivero OA, Beland FA, Poirier MC. Immunofluorescent localization and quantitation of
38-azido-28,38-dideoxythymidine (AZT) incorporated into chromosomal DNA of human,
hamster and mouse cell lines. Int J Oncol 1994;
4:49–54.
(23) Lu LJ, Anderson LM, Jones AB, Moskal TJ,
Salazar JJ, Hokanson JA, et al. Persistence,
gestation stage-dependent formation and interrelationship of benzo[a]pyrene-induced DNA
adducts in mothers, placentae and fetuses of
Erythrocebus patas monkeys. Carcinogenesis
1993;14:1805–13.
(24) Gomez DE, Kassim A, Olivero OA. Preferential incorporation of 38-azido-28,38-dideoxythymidine (AZT) in telomeric sequences of
CHO cells. Int J Oncol 1995;7:1057–60.
(25) Maniatis TE, Fritsch EF, Sambrook J. Molecular cloning: a laboratory manual. Cold Spring
Harbor (NY): Cold Spring Harbor Laboratory
Press, 1982.
(26) Percy DH, Jonas AM. Incidence of spontaneous tumors in CD (R) -1 HaM-ICR mice. J Natl
Cancer Inst 1971;46:1045–65.
(27) Kim NW, Piatyszek MA, Prowse KR, Harley
CB, West MD, Ho PL, et al. Specific association of human telomerase activity with immortal cells and cancer. Science 1994;266:2011–5.
(28) Gomez DE, Olivero OA. Long-term AZT (38azido-28,38-dideoxythymidine) treatment
1608 REPORTS
(29)
(30)
(31)
(32)
(33)
(34)
(35)
shortens the telomeric sequences of HeLa cells
in culture [abstract]. Proc Am Assoc Cancer
Res 1996;36:abstract 3848.
Anderson LM, Hagiwara A, Kovatch RM,
Rehm S, Rice JM. Transplacental initiation of
liver, lung, neurogenic, and connective tissue
tumors by N-nitroso compounds in mice. Fundam Appl Toxicol 1989;12:604–20.
Anderson LM, Ruskie S, Carter J, Pittinger S,
Kovatch RM, Riggs CW. Fetal mouse susceptibility to transplacental carcinogenesis: differential influence of Ah receptor phenotype on
effects of 3-methylcholanthrene, 7,12dimethylbenz[a]anthracene, and benzo[a]pyrene. Pharmacogenetics 1995;5:364–72.
Anderson LM, Jones AB, Riggs CW, Ohshima
M. Fetal mouse susceptibility to transplacental
lung and liver carcinogenesis by 3-methylcholanthrene: positive correlation with responsiveness to inducers of aromatic hydrocarbon metabolism. Carcinogenesis 1985;6:1389–93.
Anderson LM, Hecht SS, Dixon DE, Dove LF,
Kovatch RM, Amin S, et al. Evaluation of the
transplacental tumorigenicity of the tobaccospecific carcinogen 4-(methylnitrosamino)-1(3-pyridyl)-1-butanone in mice. Cancer Res
1989;49:3770–5.
Freireich EJ, Gehan EA, Rall DP, Schmidt LH,
Skipper HE. Quantitative comparison of toxicity of anticancer agents in mouse, rat, hamster,
dog, monkey and man. Cancer Chemother Rep
1966;50:219–44.
Balzarini J, Pauwels R, Baba M, Herdewijn P,
De Clercq E, Broder S, et al. The in vitro and
in vivo anti-retrovirus activity, and intracellular
metabolism of 38-azido-28,38-dideoxythymidine and 28,38-dideoxycytidine are highly dependent on the cell species. Biochem Pharmacol 1988;37:897–903.
Bilello JA, MacAuley C, Fredrickson TN, Bell
(36)
(37)
(38)
(39)
MM, McKissick C, Shapiro SG, et al. Use of
a neonatal murine retrovirus model to evaluate
the long-term efficacy and toxicity of antiviral agents. Ann N Y Acad Sci 1990;616:
238–51.
Bilello JA, Eiseman JL, Standiford HC, Drusano GL. Impact of dosing schedule upon suppression of a retrovirus in a murine model of
AIDS encephalopathy. Antimicrob Agents
Chemother 1994;38:628–31.
Ayers KM, Torrey CE, Reynolds DJ. A transplacental carcinogenicity bioassay in CD-1
mice with Zidovudine. Fundam Appl Toxicol
1997;38:195–8.
Yuspa SH, Poirier MC. Chemical carcinogenesis: from animal models to molecular models
in one decade. Adv Cancer Res 1988;50:
25–70.
Greenwald P, Malone WF, Cerny ME, Stern
HR. Cancer prevention research trials. Adv
Cancer Res 1993;61:1–23.
Notes
We extend our appreciation to Drs. Jerrold Ward
and Miriam Anver for review of the pathology slides
and to Dr. Robert Tarone for verifying the statistics.
For review of the data and the manuscript, we thank
Drs. Richard D. Klausner, George Vande Woude,
Alan S. Rabson, Henry C. Pitot, Lorenzo Tomatis,
and Paul Kleihues. For editorial assistance, we thank
Ms. Margaret Taylor.
Manuscript received March 7, 1997; revised August 20, 1997; accepted August 28, 1997.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
High Telomerase Activity in
Primary Lung Cancers:
Association With Increased Cell
Proliferation Rates and
Advanced Pathologic Stage
Juan Albanell, Fulvio Lonardo,
Valerie Rusch, Monika
Engelhardt, John Langenfeld, Wei
Han, David Klimstra, Ennapadam
Venkatraman, Malcolm A. S.
Moore, Ethan Dmitrovsky*
Background: Telomerase enzyme activity is not detected in most normal cells,
a phenomenon believed to be associated
with limitations on cellular proliferation. Since this activity is detected in
nearly all human tumors, including
non-small-cell lung cancers, it has been
suggested that telomerase activation
may be coupled to acquisition of the
malignant phenotype. In this study, we
determined whether telomerase activity was associated with tumor pathologic stage, tumor cell proliferation rates,
and clinical outcome in a cohort of patients with resected non-small-cell lung
cancer for whom long-term follow-up
was available. Methods: Primary tumor
specimens from 99 patients treated
with surgery alone and six patients
treated with surgery after chemotherapy were analyzed. Telomerase activity was measured by means of a
modified Telomeric Repeat Amplification Protocol (TRAP) assay. Southern
blot analysis of terminal restriction
fragments was used to evaluate telomere length. Immunohistochemical
analysis of Ki-67, a proliferationassociated nuclear antigen, was used to
assess tumor cell proliferation. Results:
Telomerase activity was detected in 84
of the 99 tumors treated with surgery
alone; this activity was not detected in
specimens of adjacent, benign lung tissue. Telomerase was detected in only
three of six tumors resected after chemotherapy. For the surgery-alone
group, statistically significant positive
associations were found between the
level of telomerase activity and tumor
stage, lymph node metastasis, pathologic TNM (tumor–node–metastasis)
stage, and Ki-67 immunostaining; a
statistically significant inverse association was found between telomerase activity and patient age. No statistically
significant differences in telomere
length were found in relation to telomerase activity or pathologic stage.
Telomerase activity was not found to be
associated with clinical outcome in a
multivariate Cox proportional hazards
analysis adjusted for tumor stage and
lymph node status. Conclusions: High
telomerase activity is detected frequently in primary non-small-cell lung
cancers that exhibit high tumor cell
proliferation rates and advanced
pathologic stage. [J Natl Cancer Inst
1997;89:1609–15]
Human telomeres are specialized nucleoprotein structures located at the ends
of chromosomes and are composed of
tandem repeats of the sequence 58TTAGGG-38 bound to specific proteins.
Conventional DNA polymerases cannot
replicate the ends of linear chromosomes,
resulting in gradual telomere shortening
when cells divide (1). Telomerase is a ribonucleoprotein that synthesizes de novo
telomeric DNA onto chromosome ends,
thus compensating for this ‘‘endreplication problem’’ (2,3). In somatic
cells, where telomerase activity is usually
not detected, there is progressive telomere
shortening during replication, and telomere length often reflects cellular proliferative potential (4–7). In contrast, germ
cells and most immortalized human cell
lines exhibit telomerase activity and
stable telomere length (8–10). Telomerase
activity is frequently expressed in human
tumors, as assessed by the highly sensitive polymerase chain reaction (PCR)based Telomeric Repeat Amplification
Protocol (TRAP) assay (8). These findings, coupled with infrequent telomerase
expression in normal cells, suggested that
telomerase activation was tightly coupled
to the acquisition of the malignant phenotype (8,11,12).
The RNA component of human telomerase (13) and a telomerase-associated
protein have recently been cloned (14).
These two components are widely expressed in human tumors and tumor cell
lines, and they may be necessary for telo-
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
mere elongation (13,14). In support of
this possibility, treatment of immortalized
human tumor cell lines with antisense oligodeoxyribonucleotides targeted to the
RNA component of human telomerase led
to a reinitiation of telomere shortening
and induction of the massive cell death
associated with proliferative senescence
(13). The maintenance of a constant average telomere length in cells expressing
telomerase activity seems to be regulated
by a negative feedback loop. This idea is
supported by functional studies with a recently cloned telomere-binding protein
that suppresses telomere elongation and is
proposed to be a negative regulator of
telomerase activity (15).
Lung cancer remains the most common cause of cancer-related death in the
United States for both men and women. A
better understanding of the biology of
lung tumors and identification of diagnostic markers or new therapeutic targets are
urgently needed to develop novel treatment strategies that could improve the
dismal survival rates of most patients with
lung cancer (16,17). In a pivotal study
(18), telomerase activity was found in
78.4% of non-small-cell lung cancers
(NSCLCs) and in all small-cell lung cancers examined. Varying levels of telomerase activity were previously detected in
NSCLCs (18), and it was hypothesized
that the tumors with no or low telomerase
activity might be composed primarily of
mortal cancer cells. High levels of telomerase activity were detected in metastatic
lesions even when undetected in the primary tumor, suggesting that telomerase
activation contributes to the development
*Affiliations of authors: J. Albanell, M. Engelhardt, W. Han, M. A. S. Moore (Laboratory of Developmental Hematopoiesis), F. Lonardo (Department of Pathology and Laboratory of Molecular
Medicine), V. Rusch, J. Langenfeld (Thoracic Surgery Service, Department of Surgery, and Laboratory of Molecular Medicine), D. Klimstra (Department of Pathology), E. Venkatraman (Biostatistics
Service, Department of Epidemiology and Biostatistics), E. Dmitrovsky (Division of Solid Tumor Oncology, Department of Medicine, and Laboratory of
Molecular Medicine), Memorial Hospital and SloanKettering Institute, Memorial Sloan-Kettering Cancer Center, New York, NY.
Correspondence to: Ethan Dmitrovsky, M.D.,
Memorial Sloan-Kettering Cancer Center, 1275
York Ave., New York, NY 10021.
See ‘‘Notes’’ following ‘‘References.’’
© Oxford University Press
REPORTS 1609
of metastatic disease. This study (18)
raised the prospect that telomerase might
represent a marker for immortal lung cancer cells and a therapeutic target in lung
cancer.
In our study, telomerase activity was
measured by use of a recently modified
TRAP assay (Kim NW, Wu F: personal
communication) that increases the reliability of the assay, allows the expression
of relative levels of telomerase activity,
and identifies the presence of inhibitors of
Taq polymerase. The purpose of this
study was to analyze comprehensively
whether a relationship existed between
telomerase activity, tumor pathologic
stage, tumor cell proliferation rates, and
clinical outcome in a well-characterized
cohort of patients with resected NSCLC
for whom long-term clinical follow-up
was available.
Methods
Tissue specimens. One hundred seven tissue
specimens were obtained from consecutive patients
with primary NSCLC who underwent potentially curative operations. Use of these found tissue specimens was approved by the relevant institutional review boards. Protein extracts from 105 of these
specimens were adequate for analysis. Pulmonary
resection was accompanied by careful intraoperative
staging with complete mediastinal lymph node dissection as described by Martini and Flehinger (19).
Lymph nodes were labeled separately for pathologic
analysis, in conformity with the American Thoracic
Society Lymph Node Map (20). Within 10 minutes
of completion of the pulmonary resection, a sample
of the primary tumor, trimmed of the surrounding
lung tissue and of any grossly necrotic material, was
snap frozen in liquid nitrogen. A separate specimen
of benign lung tissue was harvested from an area
distant from the primary tumor and also frozen in
liquid nitrogen. The presence of malignant or benign
lung tissue in the harvested samples was confirmed
by microscopic examination. An independent assessment by two pathologists (F. Lonardo and D.
Klimstra) was performed, and both pathologists concurred with respect to the histopathologic features
evaluated. The specimens were scored for tumor
grade, proportion of tumor tissue present, proportion
of necrosis in the tumors, and the estimated percentage of infiltrating lymphocytic cells in areas with
tumor. The tumor–node–metastasis (TNM) stage
was determined according to the International Staging System for NSCLC (21).
Protein extraction. Frozen tissue samples (50–
100 mg) were homogenized in 100–200 mL ice-cold
CHAPS (3-[{3-cholamidopropyl}-dimethylammonio]-1-propane-sulfonate) lysis buffer (0.5%
CHAPS, 10 mM Tris–HCl [pH 7.5], 1 mM MgCl2, 1
mM EGTA, 10% glycerol, 5 mM b-mercaptoethanol, and 10 ng/mL leupeptine) with the use of disposable pestles and standard techniques. The tissue
homogenates were incubated on ice for 30 minutes
and then centrifuged at 12 000g for 30 minutes at
1610 REPORTS
4 °C. Supernatants were collected and stored at
−80 °C. Protein concentrations were measured by
use of the Bio-Rad protein assay kit (Bio-Rad Laboratories, Richmond, CA), and aliquots containing 1
mg protein/mL were stored at −80 °C (8,22,23).
TRAP assay. The telomerase assay was performed according to a recently modified TRAP assay protocol (Kim NW, Wu F: personal communication) (8,22–24). Extracts containing 2 mg of
protein were assayed in reaction tubes that contained
50 mL of the TRAP reaction mixture. The TS primer
was labeled at its 58 end by use of 5 U T4 polynucleotide kinase (PNK; Promega Corp., Madison,WI) and 2.5 mCi of 3000 Ci/mmol [g32
P]adenosine triphosphate per 1 mg of TS. The
kinase reaction mixture was incubated at 37 °C for
20 minutes, and then the PNK was heat inactivated
at 95 °C for 5 minutes. Each TRAP reaction consisted of 5 mL 10× TRAP buffer (22), 50 mM standard deoxyribonucleoside triphosphates, 0.1 mg endlabeled TS primer, 0.1 mg RP return primer, 0.1 mg
NT internal control primer, 0.01 amol of the TSNT
internal control template, 2 U Taq DNA polymerase
(AmpliTaq; The Perkin-Elmer Corp., Branchburg,
NJ), and the extract containing 2 mg protein. TSNT
is an internal control PCR template amplified by the
primers TS and NT, giving a 36-base-pair (bp) product. After a 30-minute incubation at room temperature, the TRAP reaction mixture was subjected to 30
cycles of PCR. The PCR products were resolved by
electrophoresis in a 15% polyacrylamide gel under
nondenaturing conditions, and the gel was analyzed
on a Phosphorimager (Molecular Dynamics, Sunnyvale, CA). In every gel, the products of a negative
control reaction (2 mL CHAPS lysis buffer) and of
0.1 amol of the quantitation standard oligonucleotide R8 were included. The telomerase quantitation
results were expressed as total product generated
(TPG) (see Fig. 1 for details of the quantitation). The
specificity of the 6-bp ladders was confirmed by the
absence of the ladders following heat inactivation of
the protein extracts. All protein extracts were analyzed in at least two independent TRAP assays, and
the average telomerase activity (TPG) was calculated (Fig. 1). Subgroups were defined as having
negative (TPG 4 0), low (TPG>0 and ø5), moderate (TPG>5 and ø30), or high (TPG>30) telomerase activity.
Alkaline phosphatase activity. Alkaline phosphatase activity was assayed as a control for possible
protein degradation (23). Two tumor protein extracts
had no detectable alkaline phosphatase activity and
were not subsequently analyzed. Alkaline phosphatase levels were similar in extracts of normal and
malignant tissues (data not shown).
Immunohistochemistry. Ki-67, a proliferationassociated nuclear antigen that is present only in
proliferating cells (25), was assessed immunohistochemically to measure tumor cell proliferation rates.
Five-micron-thick, paraffin-embedded sections were
deparaffinized and rehydrated by use of standard
techniques. Pretreatment of the sections consisted of
digestion with 0.05% trypsin, followed by microwave treatment for 10 minutes. The sections were
then exposed to 3% H2O2 for 5 minutes, saturated
with 0.05% bovine serum albumin (BSA), and preincubated with normal horse serum (Cappel Research, Durham, NC) at a 1:20 dilution in 2% BSA–
Fig. 1. Telomerase activity was
measured by use of a recently
modified Telomeric Repeat
Amplification Protocol (TRAP)
assay (see text for details).
Panel A displays the R8 quantitation standard (lane 1), negative control results (Buffer [no
extract], lane 2), and the TRAP
products generated from extracts of selected pairs of nonsmall-cell lung carcinoma and
histologically benign lung tissue (lanes 4–11). The R8 quantitation standard oligonucleotide exhibits a characteristic
pattern of six bands corresponding to the first through
sixth TRAP products. The assay incorporates an internal
polymerase chain reaction
(PCR) control, yielding a 36base-pair (bp) product (designated TSNT), which migrates in the analytic polyacrylamide gel at a position
14 bp below the smallest TRAP band. This control is used to monitor PCR efficiency during the PCR step
of the assay. The amount of telomerase activity from a given reaction was calculated by use of the following
formula: TPG 4 [(T – B)/(CT)]/[(R8 – B)/(CR8)] × 100, where T 4 the radioactive counts from the
telomerase bands from the protein extract, B = the counts from the negative control reaction (background),
R8 4 the counts from the R8 standard (0.1 amol), CT 4 the counts from the internal control TSNT reaction
(0.01 amol) of the protein extract, and CR8 4 the counts from the TSNT reaction (0.01 amol) of the R8
standard (0.1 amol). The final quantitation was expressed as the TPG (total product generated). One unit of
TPG was defined as 0.001 amol (or 600 molecules) of TS primer extended by at least three telomeric repeats
by the telomerase present in the extract. A telomerase activity level of 1 TPG corresponds approximately to
the telomerase activity from one immortal cell (Kim NW, Wu F: personal communication). Panel B shows
that the assay was in the linear range from 0.001 amol (1 TPG) to 1 amol (1000 TPG) of R8 standard. This
range extended over three logarithms (base 10) of the target protein concentrations.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
phosphate-buffered saline (PBS) for 15 minutes at
room temperature. The MIB-1 antibody (Immunotech, Westbrook, ME), used at a 1:50 dilution in 2%
BSA–PBS, was applied at 4 °C for 16 hours (25).
The sections were then rinsed with PBS for 30 minutes, and a biotinylated anti-mouse immunoglobulin
G (Vector Laboratories, Inc., Burlingame, CA) was
applied at a 1:500 dilution in 1% PBS–BSA at room
temperature for 60 minutes. The sections were
rinsed with PBS and incubated with peroxidaseconjugated streptavidin (Dako Corp., Carpinteria,
CA) at a 1:500 dilution in 1% BSA–PBS at room
temperature for 45 minutes. The sections were then
rinsed with PBS for 30 minutes, and color from the
chromogen diaminobenzidine (0.06% in PBS) was
developed for 15 minutes. The sections were subsequently rinsed in water, counterstained with Harrismodified hematoxylin (Fisher Scientific Co., Pittsburgh, PA), rinsed in 1% acid alcohol for 2 seconds,
rinsed in ammonia water for 15 seconds, dehydrated,
and placed under coverslips in permount media. Ki67 immunostaining was scored as the percentage of
positive tumor cells per section.
Terminal restriction fragment (TRF) length
measurements. For DNA isolation, tissues were incubated for at least 3 hours at 50 °C with an appropriate volume of DNA extraction buffer (100 mM
NaCl, 40 mM Tris [pH 8.0], 20 mM EDTA [pH 8.0],
0.5% sodium dodecyl sulfate, and 0.1 mg/mL proteinase K), followed by phenol–chloroform–isoamyl
alcohol extractions and precipitation with 3 M sodium acetate and ethanol. Electrophoresis of the undigested, high-molecular-weight DNA was performed to assess DNA degradation. Southern blot
analysis to estimate telomere length was based on
previously reported methods (4–7,26). Genomic
DNA was digested with Msp I and Rsa I restriction
endonucleases (Boehringer Mannheim, Mannheim,
Germany) at 37 °C for an appropriate length of time.
Completeness of the DNA digestion was confirmed
by means of gel electrophoresis; 10 mg of digested
DNA was subjected to electrophoresis in 0.5% agarose gels. The resolved DNA was depurinated for 20
minutes in 0.2 M HCl, denatured for 30 minutes in
0.5 M NaOH–1.5 M NaCl, and neutralized for 30
minutes in 0.5 M Tris (pH 8.0)–1.5 M NaCl. The
DNA was then blotted overnight onto nylon filters
(Schleicher and Schuell, Inc., Keene, NH) in 20×
solution of sodium chloride and sodium citrate
(SSC). The filters were dried at 80 °C for 1 hour and
subsequently hybridized to a 32PO4 end-labeled
(TTAGGG)3 probe (Genset, La Jolla, CA) in a mixture that contained 5× SSC, 5× Denhardt’s solution,
10 mM phosphate buffer (pH 6.4), and 30 mg/mL
salmon sperm DNA at 50 °C overnight. The filters
were washed twice in 0.5× SSC–0.1% sodium dodecyl sulfate for 15 minutes each at 50 °C and then
at room temperature for an appropriate length of
time. To determine TRF length, the hybridized
probe was visualized with a Phosphorimager (Molecular Dynamics), which quantified the radioactive
signal in each of the lanes. Each lane was then
graphically divided over the range of 2–23 kilobase
pairs (kbp) into quadrants, and the densitometric
counts in each quadrant were measured. The
molecular-weight range of each quadrant was determined by use of radioactive markers. The mean
and peak TRF lengths were calculated as described
(26).
Clinical database. After pulmonary resection, the
patients were seen in follow-up by one surgeon (V.
Rusch), as previously reported (16). The parameters
that were recorded included the patient’s age and
sex, the tumor histology and stage, the estimated
percentage of viable tumor cells present in the specimen, and the disease-free and overall survivals as
calculated from the date of surgery. The telomerase
activity, TRF length, and Ki-67 immunostaining results were evaluated without knowledge of the clinical outcomes.
Statistical analysis. Associations between telomerase activity (TPG) and the sex of the patient, tumor
histology, lymph node metastasis, T stage, N stage,
and pathologic TNM stage were evaluated by use of
the Kruskal–Wallis test (27). Spearman’s rank correlations (27) were determined between telomerase
activity and patient age, Ki-67 immunostaining, percentage of tumor cells in the specimen, percentage
of necrosis in the specimen, and percentage of lymphocytic infiltration. The distribution of telomerase
activity in relation to the degree of tumor cell differentiation was compared by use of the Wilcoxon
test (27). Overall survival and disease-free survival
were calculated from the date of thoracotomy by use
of the method of Kaplan and Meier (28). The logrank test (29) was used to compare survival across
the subgroups of telomerase activity. All statistical
tests were conducted at the two-sided, .05 level of
significance. Proportional hazards regression was
used to test the prognostic significance of factors in
a multivariate model (30).
Results
Telomerase Activity and
Histopathology
A total of 105 NSCLC specimens from
105 patients were examined in this study.
Histologically benign lung tissue adjacent
to 34 of these NSCLC specimens was also
examined. Ninety-eight patients with
stage I-IIIA tumors and one patient with
stage IV disease were treated by surgical
resection alone (Table 1); six patients had
preoperative chemotherapy. The patients
were treated during the period from January 1990 through May 1996, and clinical
follow-up was updated as of September
1996. Telomerase activity was present in
84 (84.8%) of the 99 tumors from patients
treated with surgical resection alone.
Among these 84 tumors, 31 had low, 32
had moderate, and 21 had high telomerase
activity (Fig. 1; data not shown). The linearity of the TRAP assay was confirmed
over a three-logarithm (base 10) range of
the target protein concentrations (Fig. 1,
B). The average level of telomerase activity was 20 TPG (range, 0–134.9) (Fig. 2,
A). Telomerase activity was not detected
in 15 (15.2%) of the 99 tumors in addition
to all adjacent, histologically benign lung
tissue specimens examined (Fig. 1, A;
data not shown). A histopathologic review of 81 of the tumors established that
telomerase activity and the estimated percentage of tumor cells present in the
specimen (P 4 .1), the proportion of necrosis in the specimen (P 4 .72), the degree of lymphocytic infiltration (P 4 .1),
or the histologic grade of the tumor (P 4
Table 1. Comparison of tumor histology, primary tumor size, lymph node metastasis, and tumor stage
with telomerase activity in primary, resected non-small-cell lung cancer*
Telomerase activity
No. of tumors
Negative
Positive
% positive
Histology
Adenocarcinoma
Squamous cell carcinoma
Large-cell carcinoma
56
36
7
8
6
1
48
30
6
85.7
83.3
85.7
Tumor stage
T1
T2
T3
20
66
13
1
13
1
19
53
12
95.0
80.3
92.3
Lymph node status
N0
N1
N2
63
20
16
13
1
1
50
19
15
79.4
95.0
93.8
Pathologic stage
I
II
IIIA
IV†
55
18
25
1
12
1
2
0
43
17
23
1
78.2
94.4
92.0
100.0
Total
99
15
84
84.8
*See (21) for information on tumor staging.
†The corresponding patient had a simultaneous solitary brain metastasis that was resected prior to thoracotomy.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
REPORTS 1611
of MIB-1 immunostaining tumor cells
was 22%; for the telomerase-positive tumors, this percentage was 32%, 32%, and
49% for cases with low, moderate, and
high telomerase activity, respectively.
Telomerase Activity and Pathologic
Stage
Fig. 2. A) Total product generated (TPG), as a measure of the level of telomerase activity in primary,
non-small-cell lung cancers, shows the continuum of the activity detected. B) The proliferative cell fraction
was assessed by means of immunohistochemistry, using the antibody MIB-1, which recognizes the Ki-67
proliferation-associated nuclear antigen in paraffin-embedded tissue specimens. The percentage of Ki-67
staining tumor cells was correlated with telomerase activity (TPG) (r 4 +.29; P<.01). See text for additional
details.
.62) were not linked. Telomerase activity
was not associated with the sex of the
patient (P 4 .47) or the histologic tumor
type (P 4 .87) (Table 1). Patient age was
correlated inversely with telomerase activity (r 4 –.4; P<.01). The remaining six
patients had stage IIIA tumors that were
treated by surgical resection after induction chemotherapy; the data from these
patients were analyzed separately.
Telomerase Activity and Tumor Cell
Proliferation Rate
The MIB-1 antibody and immunostaining were used to measure the tumor
cell proliferation rates in 93 tumors. The
percentage of Ki-67-positive tumor cells
correlated with telomerase activity (r 4
+.29; P<.01; Fig. 2, B). In telomerasenegative tumors, the average percentage
Fig. 3. A) Telomerase activity (average total product generated [TPG] ± standard
error of the mean [SEM]) was associated with tumor size and extension of the
primary lesion (left panel, two-sided P 4 .03), the presence of lymph node
metastasis (middle panel, two-sided P 4 .05), and pathologic TNM stage (21)
(right panel, two-sided P 4 .01) in non-small-cell lung cancer. B) The Southern
blot analysis of terminal restriction fragments (TRFs) in lung tumors. The tumor
in lane 1 showed a long TRF (approximately 23 kilobases [kb]) and had high
telomerase activity; DNA from matched, benign lung tissue was unavailable for
1612 REPORTS
Telomerase activity was detected in
95.0% of T1, 80.3% of T2, and 92.3% of
T3 tumors (Table 1). The average telomerase activity (TPG) was 18 in T1 and T2
tumors compared with 36 in T3 tumors (P
4 .03) (Fig. 3, A, left panel). A significant association was found between
telomerase activity and lymph node metastasis (N0 versus N1–2, P 4 .05; Fig. 3,
A, middle panel). Telomerase activity
was detected in 79.4% of N0, 95.0% of
N1, and 93.8% of N2 tumors (Table 1).
The average telomerase activity (TPG)
was 18.3 in N0 lesions, 16.2 in N1 lesions, and 32.7 in N2 lesions. Telomerase
activity was associated with tumor pathologic stage, with average telomerase activities (TPGs) of 14.8, 18.2, and 30.2 in
stages I, II, and IIIA, respectively (P 4
.01 stage I versus stage II versus stage
IIIA; P<.01 for stage I versus stages II–
IIIA; Fig. 3, A, right panel). Telomerase
activity was detected in 78.2% of stage I,
94.4% of stage II, and 92.0% of stage IIIA
this specimen. TRFs from representative tumor (T) and matched, histologically
benign (N) lung tissue specimens are shown in lanes 2–13. Tumor mean TRF
lengths were found to be similar to mean TRF lengths in matched benign lung
tissue (compare lanes 2, 6, and 8 with lanes 3, 7, and 9, respectively), longer than
mean TRF lengths in matched benign lung tissue (compare lane 4 with lane 5 and
lane 10 with lane 11), or shorter than mean TRF lengths in matched benign lung
tissue (compare lane 12 with lane 13). The sizes of molecular weight standards
appear at the right of this figure. See text for additional details.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
tumors (Table 1). One patient with stage
IV (simultaneous solitary brain metastasis
that was resected prior to thoracotomy)
had a TPG of 108.
Telomerase Activity and Telomere
Length
The mean TRF length ranged from 6.2
to 10.4 kbp in the 15 telomerase-negative
tumors and from 5.5 to 23.0 kbp in the 30
telomerase-positive tumors examined
(Fig. 3, B; data not shown). In telomerasepositive tumors, average mean TRFs were
7.5 kpb, 8.2 kbp, and 8.2 kbp in specimens with low, moderate, and high telomerase activity, respectively. No significant
association between TRF length and tumor size, lymph node metastasis, or
pathologic stage was found. In 34 cases,
TRF analysis was performed in both the
tumor and adjacent, histologically benign
lung tissue. Similar mean TRFs were
measured in 26 (76%) cases, but TRFs
were reduced in the tumor in six (18%)
cases and elongated in the tumor in two
(6%) cases. Peak TRF values were similar
between the tumor and benign lung tissue
in 23 (68%) cases but were reduced in the
tumor in eight (24%) cases and elongated
in the tumor in three (9%) cases. Among
the four telomerase-negative tumors, the
TRF lengths were similar to those found
in normal tissues in three cases and reduced in the tumor in one case. The average telomerase activities (TPGs) were
16, 13, and 18 in tumors with elongated,
reduced, and unchanged TRF lengths, respectively.
Telomerase Activity and Clinical
Outcome
There were no statistically significant
differences in disease-free survival or
overall survival between patients grouped
on the basis of telomerase activity (P 4
.1; data not shown). The 3-year actuarial
disease-free survival was 57% for patients
with telomerase-negative tumors compared with 40% for patients whose tumors
had high telomerase activity, but this difference was not statistically significant.
The prognostic significance of telomerase activity was measured by use of a
Cox proportional hazards model. Since T
status and N status are known prognostic
factors in lung cancer, these parameters
were included in the analysis. The hazards
ratios (and 95% confidence intervals) for
overall survival were 1.78 (1.23–2.56) for
Tø2 versus T>2, 0.66 (0.48–0.91) for N
4 0 versus N>0, and 1.07 (0.63–1.82) for
telomerase-negative versus -positive
cases. Similarly, the hazard ratios (and
95% confidence intervals) for diseasefree survival were 1.77 (1.25–2.52), 0.61
(0.45–0.82), and 1.24 (0.74–2.08), respectively. No statistically significant association was found between telomerase activity and clinical outcome after this
analysis.
Telomerase Activity in Stage IIIA
NSCLC After Preoperative
Chemotherapy
Telomerase activity was not detected
in three of six primary stage IIIA tumors
resected after induction chemotherapy
(31) (data not shown). The three telomerase-negative tumors had a major pathologic response to chemotherapy. The remaining three tumors had detectable
telomerase activity and a less marked response to neoadjuvant chemotherapy.
Discussion
This study extends previous work (18)
by demonstrating that high telomerase activity in primary NSCLC is frequent in
specimens with high cellular proliferation
rates and is associated with tumors presenting with advanced stage. Telomerase
activity was detected in 84.8% of the
NSCLC specimens from patients who underwent surgical resection only. Telomerase activity was significantly higher in advanced-stage disease than in early-stage
disease. Telomerase activity was inversely associated with age, as has been
seen with at least one other tumor (32).
Telomerase activity has been linked to tumor stage in previous studies of neuroblastoma (33), breast cancer (34), gastric
cancer (35), and leukemias (36). However, such an association was not found in
studies of renal cancer (37), breast cancer
(32,38), gynecologic tumors (39), or hepatocellular carcinoma (40).
In this series, telomerase-negative tumors were infrequently associated with
lymph node metastasis at presentation.
This finding could relate to the need for
extensive cell proliferation for metastasis
to develop from individual clones. The
clonal expansion required could lead to
critically reduced telomere lengths in the
absence of telomerase reactivation, limit-
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
ing the potential for metastatic progression. In three of six stage IIIA NSCLC
cases having major pathologic responses
to chemotherapy, telomerase activity was
not detected in the tumors after chemotherapy. Perhaps repressed telomerase
activity results from effective chemotherapeutic treatment. Examining this
possibility should be the subject of future
work.
A modified PCR-based telomerase assay was used in this study. It is worth
noting that some variables could have influenced the telomerase activity measurements. For instance, alkaline phosphatase
measurements were used to assess the integrity of extracted protein. However, the
stability of alkaline phosphatase activity
may not parallel the stability of telomerase activity, which could be more sensitive to minor protein degradation in the
extracts. The efficiency of protein extraction may also have varied between the
analyzed tissue specimens. While the percentage of tumor cells present in adjacent
tissues was scored, the percentage of tumor cells present in the specimens used
for telomerase activity measurements was
not scored, since the same tissue cannot
be processed for both parameters.
A small proportion of human tumors
does not exhibit telomerase activity
(41,42). In our study, telomerase activity
was not detected in 15.2% of the primary
lung cancers treated by surgical resection
alone. In a study of retinoblastoma, a developmental tumor with a limited number
of associated mutations, telomerase activity was absent in 50% of the examined
tumors (43). It is possible that the requirement for telomerase in tumorigenesis depends on the telomere lengths found in
precursor cells and on the number of
clonal expansions needed (41). Additional
explanations for telomerase-negative tumors include the following: 1) telomerase
activation followed by its repression after
telomere elongation, 2) telomerase downregulation associated with cellular quiescence, 3) telomerase activity below the
level of detection of available assays, or
4) alternative mechanisms to compensate
for the end-replication problem. The existence of an alternative pathway was reported in immortalized cell lines and was
associated with very long telomeres (up to
50 kbp) (44). However, very long telomeres were not found in this study for the
15 telomerase-negative NSCLC speciREPORTS 1613
mens examined for telomere length. Thus
far, evidence of telomerase-negative clinical tumors having such long telomeres is
not reported.
Few studies have comprehensively addressed the relationship between telomerase activity and prognosis. In neuroblastoma, high telomerase activity correlated
with poor prognosis (33). Similar findings
are reported in gastric cancer (35) and
breast cancer (45) but not in renal cell
carcinomas (37). In this series, NSCLCs
having high telomerase activity had a
more unfavorable prognosis than telomerase-negative tumors, but the differences
were not statistically significant. While
these data suggest that telomerase has a
weak, or no, prognostic impact in lung
cancer, perhaps the addition of more patients to this series and a longer follow-up
would clarify the question of prognostic
impact of telomerase activity in NSCLC.
In this series, three patients with telomerase-negative primary NSCLC still relapsed despite surgical resection. Selection of telomerase-positive clones may
eventually occur at distant metastatic sites
when telomeres are critically shortened. It
is reported in lung cancer that telomerase
activation and telomere shortening at distant metastatic sites can occur when primary tumors are telomerase negative (18).
Since most patients with lung cancer succumb to distant disease, anti-telomerase
treatments may have a therapeutic role in
the prevention of metastasis after successful local control of the primary tumors
(9,46).
In many tumors, the mean telomere
length, as estimated by TRF analysis, is
similar to that found in the corresponding
adjacent normal tissue (33,34,41,47,48).
In this study, the mean and peak TRF
lengths in the tumors were often similar to
those in the adjacent, benign lung tissue,
although the values were occasionally either smaller or larger in the tumor. Altered TRF lengths are reported to be
linked both to high telomerase activity
(33,34) and to the lack of measurable
telomerase activity (37). In tumor-derived
cell lines, no association was found between telomere length and telomerase activity (18). As recently reported, telomere
length alone is unlikely to be an accurate
predictor of cellular immortality (34).
Telomerase activity is linked to proliferation in diverse cellular contexts (26,49).
In quiescent, primitive, hematopoietic
1614 REPORTS
progenitor cells, basal telomerase levels
are low, but the enzymatic activity is rapidly up-regulated when the cells are activated to enter the cell cycle following exposure to combinations of hematopoietic
growth factors (26). In certain human tumor cell lines, such as acute promyelocytic leukemia and human embryonal carcinoma cell lines, telomerase activity is
repressed following induced differentiation in maturation-sensitive but not maturation-resistant cell lines (22,50). Telomerase is also repressed in tumor cell lines
when the cells become quiescent, as reported previously (51,52). These observations indicate that telomerase activity and
cellular proliferation may be linked in
clinical tumors. In breast cancer, telomerase activity correlated with S-phase fraction in lymph node-positive breast cancer
(45), but other investigators (32) failed to
find such an association. In this study, a
high proliferation rate for tumor cells was
associated with telomerase activity, suggesting that telomerase is activated in
lung cancer when growth-stimulatory signals are triggered. The pattern of telomerase activity versus Ki-67 immunostaining shown in Fig. 2, B, indicates that
tumor cell subpopulations may exist,
since specimens with no or low telomerase activity were less correlated with proliferation rates than specimens with
higher telomerase activity (r 4 +.148 for
activity ø10 TPG and r 4 +.376 for activity >10 TPG, plus data not shown).
In summary, high telomerase activity
measured in primary NSCLC was found
to be associated with an increased cellular
proliferation index and advanced tumor
stage. These findings indicate that telomerase activity may contribute to lung tumorigenesis and its progression. These
observations support the concept of
telomerase as an attractive therapeutic target in lung cancer.
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
References
(1) Blackburn EH. Structure and function of telomeres. Nature 1991;350:569–73.
(2) Greider CW, Blackburn EH. Identification of a
specific telomere terminal transferase activity
in Tetrahymena extracts. Cell 1985;43:405–13.
(3) Morin GB. The human telomere terminal
transferase enzyme is a ribonucleoprotein that
synthesizes TTAGGG repeats. Cell 1989;59:
521–9.
(4) Allsopp RC, Vaziri H, Patterson C, Goldstein
S, Younglai EV, Futcher AB, et al. Telomere
length predicts replicative capacity of human
(18)
(19)
(20)
(21)
fibroblasts. Proc Natl Acad Sci U S A 1992;89:
10114–8.
Vaziri H, Dragowska W, Allsopp RC, Thomas
TE, Harley CB, Lansdorp PM. Evidence for a
mitotic clock in human hematopoietic stem
cells: loss of telomeric DNA with age. Proc
Natl Acad Sci U S A 1994;91:9857–60.
Harley CB, Futcher AB, Greider CW. Telomeres shorten during ageing of human fibroblasts. Nature 1990;345:458–60.
Hastie ND, Dempster M, Dunlop MG, Thompson AM, Green DK, Allshire RC. Telomere
reduction in human colorectal carcinoma and
with ageing. Nature 1990;346:866–8.
Kim NW, Piatyszek MA, Prowse KR, Harley
CB, West MD, Ho PLC, et al. Specific association of human telomerase activity with immortal cells and cancer. Science 1994;226:
2011–5.
Harley CB, Kim NW, Prowse KR, Weinrich
SL, Hirsch KS, West MD, et al. Telomerase,
cell immortality, and cancer. Cold Spring Harbor Symp Quant Biol 1994;59:307–15.
Counter CM, Botelho FM, Wang P, Harley
CB, Bacchetti S. Stabilization of short telomeres and telomerase activity accompany
immortalization of Epstein-Barr virus-transformed human B lymphocytes. J Virol 1994;
68:3410–4.
Counter CM, Hirte HW, Bacchetti S, Harley
CB. Telomerase activity in human ovarian carcinoma. Proc Natl Acad Sci U S A 1994;91:
2900–4.
de Lange T. Activation of telomerase in a human tumor. Proc Natl Acad Sci U S A 1994;
91:2882–5.
Feng J, Funk WD, Wang SS, Weinrich SL,
Avilion AA, Chiu C, et al. The RNA component of human telomerase. Science 1995;269:
1236–41.
Harrington L, McPhail T, Mar V, Zhou W,
Oulton R, Bass MB, et al. A mammalian
telomerase-associated protein. Science 1997;
275:973–7.
van Steensel B, de Lange T. Control of telomere length by the human telomeric protein
TRF1. Nature 1997;385:740–3.
Rusch V, Klimstra D, Venkatraman E, Pisters
P, Langenfeld J, Dmitrovsky E. Overexpression of the EGFR and its ligand, TGFa, is frequent in resectable non-small cell lung cancer,
but does not predict tumor progression. Clin
Cancer Res 1997;3:515–22.
Strauss GM, Kwiatkowski DJ, Harpole DH,
Lynch TJ, Skarin AT, Sugarbaker DJ. Molecular and pathologic markers in stage I nonsmall-cell carcinoma of the lung. J Clin Oncol
1995;13:1265–79.
Hiyama K, Hiyama E, Ishioka S, Yamakido M,
Inai K, Gazdar AF, et al. Telomerase activity in
small-cell and non-small-cell lung cancers. J
Natl Cancer Inst 1995;87:895–902.
Martini N, Flehinger BJ. The role of surgery in
N2 lung cancer. Surg Clin North Am 1987;67:
1037–49.
Tisi GM, Friedman PJ, Peters RM, Pearson
FG, Carr D, Lee RE, et al. Clinical staging of
primary lung cancer. Am Rev Respir Dis 1983;
127:659–64.
Mountain CF. A new international staging sys-
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
(22)
(23)
(24)
(25)
(26)
(27)
(28)
(29)
(30)
(31)
(32)
(33)
(34)
tem for lung cancer. Chest 1986;89(4 Suppl):
225S–233S.
Albanell J, Han W, Mellado B, Gunawardane
R, Scher HI, Dmitrovsky E, et al. Telomerase
activity is repressed during differentiation of
maturation-sensitive but not resistant human
tumor cell lines. Cancer Res 1996;56:1503–8.
Piatyszek MA, Kim NW, Weinrich SL, Hiyama K, Hiyama E, Wright WE, et al. Detection of telomerase activity in human cells and
tumors by a telomeric repeat amplification protocol (TRAP). Meth Cell Sci 1995;17:1–15.
Wright WE, Shay JW, Piatyszek MA. Modifications of a telomeric repeat amplification protocol (TRAP) result in increased reliability, linearity and sensitivity. Nucleic Acids Res 1995;
23:3794–5.
Cattoretti G, Becker MH, Key G, Duchrow M,
Schluter C, Galle J, et al. Monoclonal antibodies against recombinant parts of the Ki-67 antigen (MIB 1 and MIB 3) detect proliferating
cells in microwave-processed formalin-fixed
paraffin sections. J Pathol 1992;168:357–63.
Engelhardt M, Kumar R, Albanell J, Pettengell
R, Han W, Moore MA. Telomerase regulation,
cell cycle, and telomere stability in primitive
hematopoietic cells. Blood 1997;90:182–93.
Hollander M, Wolfe DA. Nonparametric statistical methods. New York: Wiley, 1973.
Kaplan EL, Meier O. Nonparametric estimation from incomplete observations. Am Stat
Assoc J 1958;53:457–81.
Mantel N. Evaluation of survival data and two
new rank order statistics arising in its consideration. Cancer Chemother Rep 1966;50:
163–70.
Cox DR. Regression models and life tables
[with discussion]. J R Stat Soc B 1972;34:
187–220.
Rusch V, Klimstra D, Venkatraman E, Oliver
J, Martini N, Gralla R, et al. Aberrant p53 expression predicts clinical resistance to cisplatin-based chemotherapy in locally advanced
non-small cell lung cancer. Cancer Res 1995;
55:5038–42.
Bednarek AK, Sahin A, Brenner AJ, Johnston
DA, Aldaz CM. Analysis of telomerase activity
levels in breast cancer: positive detection at the
in situ breast carcinoma stage. Clin Cancer Res
1997;3:11–6.
Hiyama E, Hiyama K, Yokoyama T, Matsuura
Y, Piatyszek MA, Shay JW. Correlating telomerase activity levels with human neuroblastoma
outcomes. Nature Med 1995;1:249–55.
Hiyama E, Gollahon L, Kataoka T, Kuroi K,
Yokoyama T, Gazdar AF, et al. Telomerase
(35)
(36)
(37)
(38)
(39)
(40)
(41)
(42)
(43)
(44)
(45)
(46)
(47)
activity in human breast tumors. J Natl Cancer
Inst 1996;88:116–22.
Hiyama E, Yokoyama T, Tatsumoto N, Hiyama
K, Imamura Y, Murakami Y, et al. Telomerase
activity in gastric cancer. Cancer Res 1995;55:
3258–62.
Counter CM, Gupta J, Harley CB, Leber B,
Bacchetti S. Telomerase activity in normal leukocytes and in hematologic malignancies.
Blood 1995;85:2315–20.
Mehle C, Piatyszek MA, Ljungberg B, Shay
JW, Roos G. Telomerase activity in human renal cell carcinoma. Oncogene 1996;13:
161–6.
Sugino T, Yoshida K, Bolodeoku J, Tahara H,
Buley I, Manek S, et al. Telomerase activity in
human breast cancer and benign breast lesions:
diagnostic applications in clinical specimens,
including fine needle aspirates. Int J Cancer
1996;69:301–6.
Kyo S, Kanaya T, Ishikawa H, Ueno H, Inoue
M. Telomerase activity in gynecological tumors. Clin Cancer Res 1996;2:2023–8.
Tahara H, Nakanishi T, Kitamoto M, Nakashio
R, Shay JW, Tahara E, et al. Telomerase activity in human liver tissues: comparison between chronic liver disease and hepatocellular
carcinomas. Cancer Res 1995;55:2734–6.
Bacchetti S, Counter CM. Telomeres and
telomerase in human cancer. Int J Oncol 1995;
7:423–32.
Shay JW, Wright WE. Telomerase activity in
human cancer. Curr Opin Oncol 1996;8:
66–71.
Gupta J, Han LP, Wang P, Gallie BL, Bacchetti S. Development of retinoblastoma in the
absence of telomerase activity. J Natl Cancer
Inst 1996;88:1152–7.
Bryan TM, Englezou A, Gupta J, Bacchetti S,
Reddel RR. Telomere elongation in immortal
human cells without detectable telomerase activity. EMBO J 1995;14:4240–8.
Kim NW, Levitt D, Huang G, Wu F, Osborne
K, Clark G. Correlation of telomerase with
prognostic indicators of breast cancer [abstract]. Proc Am Assoc Cancer Res 1996;37:
561–2.
Morin GB. Is telomerase a universal cancer
target? [editorial]. J Natl Cancer Inst 1995;87:
859–61.
Hiyama E, Yokoyama T, Hiyama K, Yamakido M, Santo T, Kodama T, et al. Alteration of telomeric repeat length in adult and
childhood solid neoplasias. Int J Oncol 1995;
6:13–6.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
(48) Hiyama K, Ishioka S, Shirotani Y, Inai K, Hiyama E, Murakami I, et al. Alterations in telomeric repeat length in lung cancer are associated with loss of heterozygosity in p53 and Rb.
Oncogene 1995;10:937–44.
(49) Hiyama K, Hirai Y, Kyoizumi S, Akiyama M,
Hiyama E, Piatyszek MA, et al. Activation of
telomerase in human lymphocytes and hematopoietic progenitor cells. J Immunol 1995;
155:3711–5.
(50) Sharma HW, Sokoloski JA, Perez JR, Maltese
JY, Sartorelli AC, Stein CA, et al. Differentiation of immortal cells inhibits telomerase
activity. Proc Natl Acad Sci U S A 1995;92:
12343–6.
(51) Holt SE, Wright WE, Shay JW. Regulation of
telomerase activity in immortal cell lines. Mol
Cell Biol 1996;16:2932–9.
(52) Zhu X, Kumar R, Mandal M, Sharma N,
Sharma HW, Dhingra U, et al. Cell cycledependent modulation of telomerase activity in
tumor cells. Proc Natl Acad Sci U S A 1996;
93:6091–5.
Notes
Present address: J. Albanell, Oncology Service,
Hospital Vall d’Hebron, Barcelona, Spain.
Supported by Public Health Service (PHS) grant
U19CA67842-01 (M. A. S. Moore) and PHS training grants T32CA09512 and K12CA01712 (J.
Langenfeld) from the National Cancer Institute, National Institutes of Health, Department of Health and
Human Services; by the Gar Reichman Fund of the
Cancer Research Institute (M. A. S. Moore); by
FIS BAE96/5706 from the Fondo de Investigacion
Sanitaria, Spain (J. Albanell); by CIRIT
1996BEAI200087 from the Comissio Interdepartmental per a la Recerca i Innovacio Technologica,
Spain (J. Albanell); and by DFG 95/3191/1-1 from
Deutsche Forschungsgemeinschaft, Germany (M.
Engelhardt). Also supported in part by the Byrne
Fund (M. A. S. Moore and E. Dmitrovsky) and
the Oracle Chemoprevention Research Fund (E.
Dmitrovsky).
We thank Dr. N. W. Kim (Geron Corporation,
Menlo Park, CA) for providing the modified TRAP
assay protocol prior to publication, Dr. K. MacKenzie (Memorial Sloan-Kettering Cancer Center, New
York, NY) for her insightful discussions, Mr. Barry
J. Nevins for expert assistance in the preparation of
the manuscript, and Ms. Melody Owens for her expert assistance in data management.
Manuscript received March 20, 1997; revised
June 6, 1997; accepted September 5, 1997.
REPORTS 1615
Interferon Alfa Versus
Chemotherapy for Chronic
Myeloid Leukemia: a
Meta-analysis of Seven
Randomized Trials
Chronic Myeloid Leukemia
Trialists’ Collaborative Group*
Background: Several randomized clinical trials in chronic myeloid leukemia
(CML) have reported better patient
survival with interferon alfa (IFN a)
than with standard chemotherapeutic
agents, such as busulfan or hydroxyurea. However, the size and persistence
of this survival benefit is uncertain.
Our aim was to assess these reliably,
both overall and in particular patient
subgroups. Methods: We collaborated
in a worldwide overview of all clinical
trials in which patients with CML were
randomly assigned to receive either
IFN a as the main drug or standard
chemotherapy. Trials were identified
by electronic and hand searching of the
medical literature and databases and
by personal contact. Individual patient
data were available for each of 1554
patients who had been randomly assigned to treatment in seven trials
(German, Italian, British, French,
Japanese, and ‘‘Benelux’’). Intentionto-treat stratified logrank survival
analyses were performed, reporting
two-sided P values. Results: Almost all
of the patients in these trials had disease with the Philadelphia chromosome
abnormality. Among those who did, the
regimens that involved IFN a produced
a statistically significantly better survival than those involving either hydroxyurea (P = .001) or busulfan (P =
.00007) alone. The 5-year survival rates
were 57% with IFN a and 42% with
chemotherapy, with an absolute difference of 15% (standard deviation = 3%;
P<.00001). There were no trials or subgroups of patients in which the treatment difference was statistically significantly different from the average.
Conclusion: For patients with Philadelphia chromosome-positive chronic myeloid leukemia, the inclusion of IFN a
1616 REPORTS
in the therapeutic regimen produced
substantially better 5-year survival
than standard chemotherapy alone. [J
Natl Cancer Inst 1997;89:1616–20]
Chronic myeloid leukemia (CML) is a
hematopoietic stem cell disorder that generally progresses, after some years of
treatment, from a relatively benign
chronic phase to an acute aggressive stage
(blast crisis). Most patients diagnosed as
having CML have leukemic cells with the
Philadelphia (Ph) chromosome, in which
a translocation between chromosomes 9
and 22 has resulted in the fusion of the
BCR and ABL genes (which encode a
serine-threonine kinase and a tyrosine kinase, respectively) to form an oncogene
(1). This definite chromosomal abnormality allows reduction of the population of
leukemic cells to be monitored, and interferon alfa (IFN a) can cause at least temporary disappearance of the disease, as
monitored by basic cytogenetics, in some
patients.
Since the introduction of IFN a for the
treatment of CML, several randomized
clinical trials have reported significantly
better survival for patients treated with
this biologic response modifier than for
patients treated with standard chemotherapeutic agents, such as busulfan or
hydroxyurea (2–5). To assess reliably the
size and persistence of any survival benefit and to establish whether there is a
particularly large or small benefit in various subgroups, a worldwide collaborative
overview of the randomized evidence has
been conducted. This form of analysis has
two advantages; it avoids selective overemphasis on the results of particular studies and, because large numbers are involved, it reduces the effects of the play
of chance. The aim of this collaboration
was to assess the difference in survival of
patients when IFN a is compared with
standard chemotherapy and to establish
whether any such difference is greater in
particular types of patient.
Materials and Methods
Randomized trials of CML treatment were sought
that began before 1990 and compared IFN a with
chemotherapy—that is, versus busulfan, versus hydroxyurea or, in a three-way randomization, versus
both. (These three-way trials also provided directly
randomized comparisons of busulfan versus hydroxyurea, since they involved randomization to
three treatment groups: IFN a versus busulfan ver-
sus hydroxyurea.) Medline and computerized clinical trial databases were searched; meeting abstracts,
reference lists, and review articles were examined;
and experts and pharmaceutical companies were
contacted (6). Once a relevant trial had been identified, information, including the most recent followup available, was sought on each randomized
patient. This information included sex, Ph chromosome positivity, platelet and white blood cell counts,
and Sokal score at diagnosis, along with dates of
birth, diagnosis, and randomization. The follow-up
variables collected included date last seen alive, date
and type of response, date and type of any bone
marrow transplant, and date and cause of death.
These data were checked centrally for any obvious
inconsistencies (e.g., dates out of order), for balance
between treatments within different subgroups and
over time, and for apparent discrepancies with any
publications. As far as possible, data were to be
obtained for all patients who had ever been randomized, irrespective of whether they had been included
in previously published trial analyses. Tabulations
of their own data (listing the numbers randomized
and the numbers of deaths in various subgroups)
were sent to the trialists for checking, thereby ensuring that the data had been correctly interpreted.
The methods for this type of meta-analysis (i.e.,
based on individual patient data) have been described in detail elsewhere (6).
Some trials allowed random assignment of Phnegative patients, but the main analyses involve only
the Ph-positive patients. Since the number of patients not known to be Ph positive was small, however, their inclusion or exclusion has no material
effect on the overall result.
Statistical Analysis
Intention-to-treat analyses were used, with patients compared on the basis of their randomly allocated treatment, regardless of whether it was actually received. Logrank survival analyses for each
trial yielded, for the IFN a-allocated group, the observed number of deaths (O), the expected number
of deaths (E), and the variance (V) of the difference
between these values for each trial (O–E). These
were then summed, one per trial, and used to calculate the death rate ratio [by the ‘‘one-step’’ approximation exp[(O–E)/V] to the hazard function ratio
(7,8)]. These death rate ratios are sometimes described in terms of percentage odds reductions: thus,
for example, a ratio of 0.70 with standard deviation
(SD) 0.06 could be described as a reduction of 30%
(SD 4 6%) in the annual death rate. All P values
quoted are two-sided. For trials that involved 2:1 (or
1:2) randomization, the logrank tests are calculated
in the usual way (7), but, to balance the contribution
from each randomization between the two treatment
arms, the chemotherapy group counts twice (or half)
just in the crude subtotals of deaths and patients.
Trials Included
Table 1 shows the 11 trials that were identified.
Data were available for seven of them. The missing
*Correspondence to: CML Trialists’ Collaborative Group, Clinical Trial Service Unit, Radcliffe
Infirmary, Oxford OX2 6HE, U.K.
See ‘‘Notes’’ following ‘‘References.’’
© Oxford University Press
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
Table 1. List of all relevant randomized trials that began before 1990: interferon alfa (IFN a) versus busulfan or hydroxyurea*
Study
(reference No.)
Induction
treatment
IFN a
dose
IFN a
target
WBC
Additional therapy
if necessary
Alternative
treatment
Chemotherapy
target WBC
Ph-negative
patients
eligible
Patients
Deaths
Median years
follow-up of
survivors
No. of
Trials that are available
ItalianCML-86 (2)
GermanCML-1 (3)
MRC-CML-3
(4)
Benelux (9)
Pessac (10)
EORTC-06887
IFN a
9 MU/d
IFN a
5 MU/m2/d
2–4
Chemotherapy
Chemotherapy
IFN a
Chemotherapy
3 MU/d
2–5
Japan (5)
IFN a
Washington (11)
Scheringplough
MEX-INCCML
CastillaLeon (12)
IFN a
3 MU × 5/w
<10
2
4 MU/m /d
5 MU/d
<10
9 MU/d
5
HU or Bu if WBC
>30
nil
HU or Bu if WBC
>30
HU to keep WBC
<10
nil
A-BMT if complete
cytogenetic
response
nil
HU
<30
No
322
198
9
Bu
HU
Bu
HU
HU
<20
5–15
4–20 or <30
4–20
<10
Yes
603
329
3
Yes
590
331
4
No
197
72
4
HU
HU
4–10
<15
Yes
Yes
22
78
10
20
6
2
Bu
5
No
170
60
4
Trials that are not available
2
4 MU/m /d
HU
18
IFN a
4 MU/m /d
<15
HU
4–10
133
IFN a
5 MU/m2/d
<20
Bu
<20
30
Chemotherapy
2 MU/d
<10
HU
5–10
26
2
HU to keep WBC
<10
*HU 4 hydroxyurea, Bu 4 busulfan, A-BMT 4 autologous bone marrow transplant, MU/m2/d 4 mega-units of interferon/square meter/day, WBC 4 white
blood cell count (in units of 109/L); note that in some trials Ph-negative patients were not eligible.
trials were small, however, so in total about 90% of
the patients in these 11 trials have been included.
In the British Medical Research Council (MRC)
CML-3 (MRC-CML-3) trial, patients were induced
with busulfan or with hydroxyurea (selected either
by randomization or, often, by physician choice) and
were randomized to receive IFN a or to continue
with the treatment that had been used for induction.
This trial was therefore split into two, one part comparing IFN a versus busulfan and the other comparing IFN a versus hydroxyurea.
The German CML-1 trial began as a 1:1 randomization between busulfan and hydroxyurea, with IFN
a introduced as a third arm for most centers in 1986
(initially in the ratio 1:1:1 and later in the ratio 1:1:
2). In this report, these three parts of the study are
analyzed separately and the results then summed.
Only the second and third parts contribute directly to
the comparison of IFN a versus busulfan, of IFN a
versus hydroxyurea, or (by logrank analyses that
avoid any double counting) of IFN a versus chemotherapy (either busulfan or hydroxyurea) (Fig. 1).
Results
IFN a Versus Chemotherapy
Fig. 1 shows the results among Phpositive patients for each separate trial.
There is an overall reduction in the annual
death rate of 30% (SD 4 6%), which is
highly statistically significant (P<.00001).
Confidence intervals for each of the trials
overlap this average reduction, indicating
that no trial had a result that was statisti-
cally significantly different from the average, and a formal test for heterogeneity
between the different trial results was
likewise not statistically significant. Some
patients received a bone marrow transplant in chronic phase, but censoring at
transplant makes no material difference to
the results, giving an overall reduction
in the annual death rate of 32% (SD 4
7%).
The overall reduction in the annual
death rate is 26% (SD 4 8%; P 4 .001)
in the trials of IFN a against hydroxyurea,
and 36% (SD 4 9%; P 4 .00007) in the
trials of IFN a against busulfan. Although
the comparison of IFN a versus busulfan
yields a somewhat greater reduction, this
result is not significantly different from
the result yielded by the comparison of
IFN a versus hydroxyurea. Fig. 2 shows
the cumulative effect on survival. Median
survival is prolonged by about 1 or 2
years and there is an improvement in the
5-year survival rates from 42% with
chemotherapy to 57% with IFN a, an
absolute difference of 15% (SD 4 3%;
with logrank P<.00001). The absolute
improvement in the 5-year survival rate
is 20% (SD 4 5%) in the trials of
IFN a versus busulfan and 12% (SD 4
4%) in the trials of IFN a versus hydroxyurea.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
For Ph-positive patients, analyses
within Sokal stage (13), age, and sex
groups did not reveal any heterogeneity
between the sizes of the effects in any of
these subgroups (Fig. 3). The analyses
among Ph-negative and Ph-unknown patients are shown separately in Fig. 3, but
there were too few such patients for these
results to be informative.
The results in different time periods
are given separately in Fig. 3. Throughout
the first 5 years, the annual death rate was
lower in those allocated IFN a than in
those allocated chemotherapy, but after 5
years it appears not to be. This does not,
however, suggest that the benefit accrued
during the first 5 years is then lost. It
merely suggests that, among those who do
survive to 5 years (which will be a greater
proportion of one group than of the other;
Fig. 2), the subsequent prognosis is about
the same in both groups. Thus, the crude
mortality during the first 5 years is 312 of
864 for the IFN a patients versus 415 of
830 for the adjusted control group (i.e.,
36% versus 50%, corresponding to an absolute difference of 14%, which is similar
to the absolute difference of 15% in the
5-year survival in Fig. 2). The numbers
still alive and being followed at the end of
the first 5 years were 254 for the IFN a
patients versus 185 for the adjusted conREPORTS 1617
but the numbers randomized between
these two chemotherapy agents are too
small for such subgroup analyses to be
reliable.
Discussion
Fig. 1. Ratios of annual death rates in the randomized trials of interferon alfa (IFN a) versus control
(hydroxyurea or busulfan) in Philadelphia chromosome-positive chronic myeloid leukemia (CML): combination of evidence from different trials. Each trial result is represented by a square, with larger squares for
trials with more data and a horizontal line indicating the 99% confidence interval (CI). Squares to the left
of the solid vertical line indicate better results with IFN a, but if the CI crosses this line, the result is not
significant at the P 4 .01 level. Subtotals and the overall result are represented by diamonds whose widths
show the 95% CIs, accompanied by the percentage odds reduction and its standard deviation (SD). Note:
Where a trial had several parts, with separate randomization procedures, each part is analyzed separately and
the results of these analyses are added together. Where randomization was in a 2:1 (or 1:2) ratio, twice (or
half) the numbers of deaths and patients in the chemotherapy arm are added into the adjusted (adj) total to
balance the contribution from each study, but this adjustment does not affect the calculation of the logrank
observed minus expected (O–E) or its variance. The German trial was in three parts in which the hydroxyurea:busulfan:IFN a allocation ratios differed, being (i) 1:1:0 (with 71 deaths of 105 to allocated hydroxyurea
and 72 deaths of 109 to allocated busulfan); (ii) 1:1:1 (with mortality 13 of 35:22 of 32:12 of 27, respectively); and (iii) 1:1:2 (with mortality 22 of 53:20 of 45:30 of 104, respectively). Although there are 510
patients in parts i, ii, and iii, only the 296 patients in parts ii and iii provide directly randomized comparisons
between IFN a and chemotherapy. (The logrank O–E and its variance for IFN a versus chemotherapy were
−0.8 and 9.3 in part ii and −6.3 and 17.9 in part iii.)
trol group. The crude death rate among
them after the first 5 years was 80 of 254
for the IFN a patients versus 55 of 185 for
the adjusted control group (31% versus
30%), indicating no further difference.
However, since 85% of the deaths thus far
reported in these trials occurred during
the first 5 years, the evidence on what
happens later is limited, as may be seen
from the wide 99% confidence interval
for the final black square in Fig. 3.
Hydroxyurea Versus Busulfan
The three-way comparison in the German CML-I trial (14) and the induction
1618 REPORTS
randomization in the MRC-CML-3 trial
(4) also provide some limited evidence on
hydroxyurea versus busulfan. In these two
trials, hydroxyurea appeared to be the better option. In comparison with allocation
to busulfan, allocation to hydroxyurea reduced the proportional odds of death by
24% (SD 4 10%; P 4 .02), suggesting
an absolute improvement in the 5-year
survival rate of about 10% but with wide
confidence intervals. Again, analyses
within Sokal stage, age, sex, and Ph status
groups did not show any statistically significant heterogeneity between the effects
in different subgroups (data not shown),
For patients with Ph-positive CML,
these seven randomized trials that compared IFN a (as the main drug) versus
continued chemotherapy demonstrate a
highly statistically significant survival
benefit for the regimens that involved IFN
a, with an absolute improvement in 5year survival of 15% (SD 4 3%) from
42% to 57%. This estimate may be somewhat reduced, due to the inclusion in
some trials of patients who were not in
early chronic phase. The majority of patients with CML are over the age of 60
years, and since repeated injections of
IFN a can involve considerable inconvenience, costs, and side effects (2,3,5), it
would be useful to know whether there
are some recognizable types of patient
who can expect little benefit. Unfortunately, however, this cannot be determined reliably from these trials.
A larger treatment effect in the Sokal
stage 1 subgroup was reported in the
MRC CML-3 trial (4), but this is not confirmed in the overview. There was no statistically significant evidence of any different treatment effect in any particular
sex, age, or risk group, although the possibility that IFN a is more beneficial in
some groups than others cannot be excluded. The number of patients randomized who were not Ph positive was too
small to be informative. It is possible that
the size of benefit obtained with IFN a
varies according to the degree and time of
hematologic and cytogenetic response.
However, the information available from
these trials is not sufficient to investigate
this properly. If chemotherapy is to be
used, hydroxyurea appears better than busulfan, but regimens that involve IFN a
result in statistically significantly better
5-year survival than those involving either chemotherapeutic agent. There was
no statistically significant heterogeneity
of treatment effect between these trials,
even though they used a variety of different treatment policies, but this does not
preclude the possibility that some policies
are better than others. In particular, it is
not yet clear what the dosage or duration
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
Fig. 2. Survival rates during the first 5 years in the randomized trials of (A) interferon alfa (IFN a) versus hydroxyurea, (B) IFN a versus busulfan, and (C) IFN a versus
control (hydroxyurea or busulfan) in Philadelphia chromosome-positive chronic myeloid leukemia. Descriptive survival curve calculated (by modified Kaplan–Meier method)
from the logrank observed minus expected and its variance
in each separate year. Annual death rates in each period are
calculated from the total number of deaths divided by the
total ‘‘person-years.’’ Numbers at risk at the start of each
year are shown for patients allocated IFN a and for the
adjusted control group. Vertical lines indicate one standard
error above or below each plotted point.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997
REPORTS 1619
(8)
(9)
(10)
(11)
(12)
(13)
(14)
Fig. 3. Ratios of annual death rates in the randomized trials of interferon alfa (IFN a) versus control
(hydroxyurea or busulfan) in chronic myeloid leukemia, subdivided by patient characteristics and years since
randomization. Ph 4 Philadelphia chromosome; SD 4 standard deviation; and O–E 4 logrank observed
minus expected. No statistically significant heterogeneity of effect was found with respect to any of these
factors.
of IFN a should be or how best to combine IFN a with chemotherapy.
References
(1) Gordon MY, Goldman JM. Cellular and molecular mechanisms in chronic myeloid leukaemia: biology and treatment. Br J Haematol
1996;95:10–20.
(2) Interferon alfa-2a as compared with conventional chemotherapy for the treatment of
chronic myeloid leukemia. The Italian Cooperative Study Group on Chronic Myeloid Leukemia. N Engl J Med 1994;330:820–5.
(3) Hehlmann R, Heimpel H, Hasford J, Kolb HJ,
Pralle H, Hossfield DK, et al. Randomized
comparison of interferon-alpha with busulfan
and hydroxyurea in chronic myelogenous leukemia. Blood 1994;84:4064–77.
(4) Allan NC, Richards SM, Shepherd PC. UK
Medical Research Council randomised, multi-
1620 REPORTS
centre trial of interferon-alfa n1 for chronic
myeloid leukaemia: improved survival irrespective of cytogenetic response. The UK
Medical Research Council’s Working Parties
for Therapeutic Trials in Adult Leukemia. Lancet 1995;345:1392–7.
(5) Ohnishi K, Ohno R, Tomonaga M, Kamada N,
Onozawa K, Kuramoto A, et al. A randomized
trial comparing interferon-alpha with busulfan
for newly diagnosed chronic myelogenous leukemia in chronic phase. Blood 1995;86:
906–16.
(6) Stewart LA, Clarke MJ. Practical methodology
of meta-analyses (overviews) using updated individual patient data. Cochrane Working
Group. Stat Med 1995;14:2057–79.
(7) Early Breast Cancer Trialists’ Collaborative
Group. Systemic treatment of early breast cancer by hormonal, cytotoxic, or immune
therapy. 133 randomised trials involving
31,000 recurrences and 24,000 deaths among
75,000 women. Early Breast Cancer Trialists’
Collaborative Group. Lancet 1992;339:
1–15.
Secondary prevention of vascular disease by prolonged antiplatelet treatment. Antiplatelet Trialists’ Collaboration. Br Med J 1988;296:320–31.
Low-dose interferon-alpha 2b combined with
hydroxyurea versus hydroxyurea alone for
chronic myelogenous leukemia. The Benelux
CML Study Group. Bone Marrow Transplant
1996;17(Suppl 3):s19–s20.
Broustet A, Reiffers J, Marit G, Fiere D, Jaubert J, Reynaud J, et al. Hydroxyurea versus
interferon alfa-2b in chronic myelogenous leukaemia: preliminary results of an open French
multicentre randomized study. Eur J Cancer
1991;27(Suppl 4):s18–s21.
Jacobson RJ, St. Germain D. Reduced toxicities using interferon-alfa (IFN) and hydroxyurea in chronic myelogenous leukemia (CML).
ECCO meeting abstracts 1991;6:s254.
Moro MJ, Gil S, Canizo C, Clemente J Fernandez, Guerras L, Fisac RM, et al. The treatment of
chronic myelogenous leukemia with interferon
alfa-2b plus hydroxyurea versus hydroxyurea
alone. Haematologica 1991;76(Suppl 4):117.
Sokal JE, Cox EB, Baccarani MI, Tura S,
Gomez GA, Robertson JE, et al. Prognostic
discrimination in ‘‘good-risk’’ chronic granulocytic leukemia. Blood 1984;63:789–99.
Hehlmann R, Heimpel H, Hasford J, Kolb HJ,
Pralle H, Hossfield DK, et al. Randomized
comparison of busulfan and hydroxyurea in
chronic myelogenous leukemia: prolongation
of survival by hydroxyurea. The German CML
Study Group. Blood 1993;82:398–407.
Notes
Supported by the Imperial Cancer Research Fund,
the Medical Research Council, and the Biomed program of the European Union (grant PL-931247).
The groups and trialists who collaborated are
listed alphabetically as follows: Benelux Chronic
Myeloid Leukemia (CML) Study Group, Belgium
and The Netherlands (A. Delannoy, J. C. KluinNelemans, and A. Louwagie); EORTC, Leukaemia
Cooperative Group (P. Stryckmanns and S. Suciu);
German CML Study Group (R. Hehlmann, H.
Heimpel, and J. Hasford); Italian Cooperative Study
Group on CML (E. Zuffa, M. Baccarani, and S.
Tura); Japan Kouseisho Leukemia Study Group (K.
Ohnishi, R. Ohno); Medical Research Council, U.K.
(N. Allan and P. Shepherd); and Pessac, France (A.
Broustet). Secretariat (R. Alison, M. Clarke, H.
Duong, R. Gray, E. Greaves, R. Peto, S. Richards,
D. Sinclair, and K. Wheatley). Writing Committee
(S. Richards, N. Allan, M. Clarke, R. Gray, and R.
Peto).
Manuscript received December 16, 1996; revised
May 29, 1997; accepted August 28, 1997.
Journal of the National Cancer Institute, Vol. 89, No. 21, November 5, 1997